Professional Documents
Culture Documents
... (2)
where n represents the number of sensor nodes
are randomly deployed, d is the distance from a
node to the cluster head, M is the sensing area of x
or y axis, and B is the distance from the centre of
sensing area to the outside location of the base
station. Moreover, this paper also contains two
phases for the entire network that are setup
phase and transmission phase. By the setup
phase, sensor nodes are organized into clusters
with the proposed fixed optimal cluster numbers.
With the proposed fixed optimal cluster numbers,
every round uses the optimal cluster number to
form clusters. During the transmission phase, the
data is transmitted back to the cluster head and
then cluster head will forward the data back to
the base station
IV. A FIXED OPTIMAL CLUSTER
NUMBER PROTOCOL
As we can see the architecture of LEACH and
HEED cannot evenly dissipate the energy of all
nodes because of the uneven clusters selection for
each round in the network Therefore, we firstly
find the optimal cluster numbers for the LEACH
architecture and then use it with the random
deployment. However, there are two phases in the
network. The first phase is setup phase including
sensor nodes randomly deployment, and k
optimal cluster number founded for each round.
The second phase is transmission phase including
TDMA and data aggregation.
Fig. 2: Fixed optimal cluster number protocol.
The following lists are two phases in the network.
1. Nodes firstly deployed into M x M region.
2. Given the B.S. location.
3. B.S. is set to the centre of the sensing area; uses
the fixed optimal cluster numbers in (1) for the
cluster formation phase.
NICE-2010
Acharya Institute of Technology, Bangalore-560090 24
4. B.S. is set to the outside of the sensing area;
uses the fixed optimal cluster numbers in
(2) for the cluster formation phase.
5. Nodes organize themselves into clusters with
the given B.S. location address.
6. Every round will have the same fixed optimal
cluster numbers.
7. Data can be sent back to the cluster head and
forward to base station with TDMA and data
aggregation.
V. SIMULATION RESULTS
In order to have better simulation result, the
simulation tool, Mat lab, is used as simulator. In
the simulation environment, 100 sensor nodes are
randomly deployed as shown in Fig. 3
Fig. 3: 100 sensor nodes are randomly deployed
Fig. 4 shows the formation of clusters and how
many nodes are still alive there. o denotes
sensor node is still alive where as . denotes node
is no longer alive and * denotes cluster head.
Also, sensor nodes colour stands for which node
belongs to which cluster.
.
Fig.4: Cluster formation and cluster head selection.
By evaluating 310 rounds in Fig. 5 and 1200
rounds in Fig. 6, numbers of cluster head from
both figures are different and nodes number
belongs to each cluster from these two figures are
also different. Therefore, by Fig. 10, it shows those
sensor nodes closer to the base station will be
runout of energy quickly.
Fig. 5: Sensor nodes are still alive after 310 rounds.
Fig. 6 Sensor nodes are still alive after 310 rounds.
As the base station is set at the centre of the
sensing area, Fig. 6 shows how many nodes are
still alive and how many nodes are no longer alive.
It is obvious that energy usage for nodes far away
from the base station is less than those nodes
closer to the base station. Therefore, the
probability and cluster formation need to be
adjusted in order to have the dissipated energy
evenly.
VI. CONCLUSION
In this paper, we have presented a distributed,
energy efficient
Clustering approach for ad hoc sensor networks.
Our approach is hybrid: Cluster heads are
probabilistically selected based on their residual
energy, and nodes join clusters such that
communication cost is minimized. We assume
quasi-stationary networks where nodes are
location-unaware and have equal significance. A
key feature of our approach is that it exploits the
availability of multiple transmission power levels
at sensor nodes and placing the base station at the
centre. Based on this approach, we have
introduced the AHEED protocol, which terminates
in a constant number of iterations, independent of
network diameter. Simulation results
demonstrate that AHEED prolongs network
lifetime, and the clusters it produces exhibit
several appealing characteristics. AHEED achieves
a connected multihop intercluster network when
a specified density model and a specified relation
between cluster range and transmission range
NICE-2010
Acharya Institute of Technology, Bangalore-560090 25
hold. Our approach can be applied to the design of
several types of sensor network protocols that
require scalability, prolonged network lifetime,
fault tolerance, and load balancing. Although we
have only provided algorithms for building a two-
level hierarchy, we can extend the protocols to
multilevel hierarchies. This can be achieved by
recursive application at upper tiers using bottom-
up cluster formation .The sensors which become
the cluster-head in LEACH architecture spend
relatively more energy than other sensors
because they have receive information from all
the sensors within their cluster, aggregate this
information and then communicate to the base
station. Hence, they run out of their energy faster
than other sensors. We have found the optimal
number of cluster-heads for the proposed
algorithm that minimize the energy spent in the
network, when sensors are uniformly distributed
in a bounded region. We know the number of
cluster-heads depend on the distance between
base station and sensor network system.
VII. BIBLIOGRAPHY
VIII.
[1] D. Estrin, L. Girod, G. Pottie, and M. Srivastava,
Instrumenting the World with Wireless Sensor
Networks, Proc. Intl Conf. Acoustics, Speech, and
Signal Processing (ICASSP 2001), May
2001,http://citeseer.nj.nec.com/estrin01instrum
enting.html.
[2] G.J. Pottie and W.J. Kaiser, Wireless Integrated
Newtork Sensors, Comm. ACM, vol. 43, no. 5, pp.
51-58, May 2000.
[3] V. Kawadia and P.R. Kumar, Power Control
and Clustering in Ad Hoc Networks, Proc. IEEE
INFOCOM, Apr. 2003.
[4] S. Narayanaswamy, V. Kawadia, R.S. Sreenivas,
and P.R. Kumar, Power Control in Ad-Hoc
Networks: Theory, Architecture, Algorithm and
Implementation of the COMPOW protocol, Proc.
European Wireless 2002. Next Generation
Wireless Networks: Technologies, Protocols,
Services and Applications, pp. 156-162, Feb. 2002.
[5] C. Intanagonwiwat, R. Govindan, and D. Estrin,
Directed Diffusion: A Scalable and Robust
Communication Paradigm for Sensor Networks,
Proc. ACM/IEEE Intl Conf. Mobile Computing and
Networking (MOBICOM), 2000.
[6] J. Kulik, W. Heinzelman, and H. Balakrishnan,
Negotiation- Based Protocols for Disseminating
Information in Wireless Sensor Networks}, ACM
Wireless Networks, vol. 8, nos. 2-3, pp. 169-185,
2002,
citeseer.nj.nec.com/kulik99negotiationbased.
html.
[7] J.-H. Chang and L. Tassiulas, Energy
Conserving Routing in Wireless Ad-Hoc
Networks, Proc. IEEE INFOCOM, Mar. 2000,
http://www.ieee-
infocom.org/2000/papers/417.ps.
[8] W. Heinzelman, A. Chandrakasan, and H.
Balakrishnan, An Application-Specific Protocol
Architecture for Wireless Microsensor Networks,
IEEE Trans. Wireless Comm., vol. 1, no. 4, pp. 660-
670, Oct. 2002.
[9] A. Cerpa and D. Estrin, ASCENT: Adaptive
Self-Configuring Sensor Networks Topologies,
Proc. IEEE INFOCOM, June 2002.
[10] Y. Xu, J. Heidemann, and D. Estrin,
Geography-Informed Energy Conservation for Ad
Hoc Routing, Proc. ACM/IEEE Intl Conf. Mobile
omputing and Networking (MOBICOM), pp. 70-84,
July 2001.
[11] B. Chen, K. Jamieson, H. Balakrishnan, and R.
Morris, Span: An Energy-Efficient Coordination
Algorithm for Topology Maintenance in Ad Hoc
Wireless Networks, ACM Wireless Networks, vol.
8, no. 5, Sept. 2002.
[12] C.R. Lin and M. Gerla, Adaptive Clustering for
Mobile Wireless Networks, IEEE J. Selected Areas
Comm., Sept. 1997.
[13] S. Banerjee and S. Khuller, A Clustering
Scheme for Hierarchical Control in Multi-Hop
Wireless Networks, Proc. IEEE INFOCOM, Apr.
2001.
[14] B. McDonald and T. Znati, Design and
Performance of a Distributed Dynamic Clustering
Algorithm for Ad-Hoc Networks, Proc. Ann.
Simulation Symp., 2001.
NICE-2010
Acharya Institute of Technology, Bangalore-560090 26
BUILDING EXTRACTION IN AERIAL IMAGES USING SCALING AND
MORPHOLOGICAL TECHNIQUES
Anjum mujawar, Prof. J.N.Ingole
Assist. Prof., Electronics Dept, Dr Meghe Institue of technology & Research,Amravati university
pro.anjum@gmail.com
_____________________________________________________________________________________________________________________________
ABSTRACT
This paper presents the applicability of
image scaling and morphological operators to the
detection of building in grey scale aerial
photography. The basic idea of scaling is to
generate a multiscale representation from one
parameter family of derived signals. The second
method involves several morphological operators
among which an adaptive hit-or-miss transform
with varying size and shapes of structuring element
intended to determine the optimal filtering
parameters automatically. We have implemented
and tested both the operations on various images.
KEYWORDS
Scaling, Homogeneity operator, Mathematical
Morphology, Hit Or Miss Transform,
Bidimensional Granulometry.
I. INTRODUCTION
A significant amount of work that has been
done in the field of aerial image understanding
has concentrated on the detection of buildings
and roads.
We examine the use of image scaling and
morphological operator. The basic idea of scale
space theory is to generate a multi scale
representation one parameter family of derived
signals. The scale parameter is intended to
describe the current level of scale. The space is
constructed by smoothing the original image with
Gaussian kernels of successively increasing width.
Morphology has proved to be effective for
many applications in remote sensing. Our method
is
based on a sequence of different morphological
operators applied on binary images among which
the Hit- or-Miss transforms We compute a
granulometry
[13] to determine the parameters of the operators
in an automatic way.
The paper is organized as follows. In
section II, we present the building extraction
method based on scaling. In section III,
morphological operator method is
explained. In section IV, conclusion is given
references are given in section V.
II. SCALING
A. Theory
Scaling an image involves the generation of a
family of images at different scales from a single
source image [].The image family is created by
convolving Gaussian kernels of successively
increasing widths with the original image. This
results in the source image becoming increasingly
smoothed and successive suppression of fine-
scale information [6]. The discrete Gaussian
kemel (h (n1, n2)) in two dimensions is given by
the following two equations.
( )
( )
2 2 2
1 2
, / 2
1 2
,
n n
g
h n n e
= (0.1)
( )
( )
1 2
1 2
,
,
g
g
h n n
h n n
h
=
(0.2)
Where is the standard deviation and (n1,n2)
indexes a specific pixel in the kernel.
Scaling an image cannot be carried out with
any kind of smoothing or averaging operation. It
is essential that the smoothing operation does not
introduce any artifacts as the transformation from
coarse to fine scales occurs. Only Gaussian
smoothing obeys this criterion
Scaling or scale-space events are linked to
abstraction. Abstraction according to Mayer [6] is
defined as the increase of the degree of
simplification and emphasis. As an aerial image is
progressively smoothed the substructure of
houses and other objects are suppressed. At high
levels of smoothing small objects are completely
annihilated and large objects appear as blobs,
having lost all substructure. According to Mayers
definition abstraction is occurring - unnecessary
information in the image has been eliminated
resulting in an emphasis of what remains.
NICE-2010
Acharya Institute of Technology, Bangalore-560090 27
Figure (1): Source Image
B. Investigation
An investigation into the effects of image
scaling has been carried out on the following
aerial image depicts a sparsely built-up area.
There are a number of houses, a few road
segments, some natural vegetation, and large
barren sand patches. This variety of scenery is
useful as it allows the effects of smoothing and
abstraction to be more accurately evaluated.
Figure 2 is a Gaussian smoothed version of Figure
1. A Gaussian kemel of size 15 by 15 pixels was
used with a standard deviation of 3 Notice how
the fine image detail is suppressed, in particular,
the road markings of the image have begun to blur
into the road itself and various cars in the image
are almost unrecognizable.
Figure 3 represents the source image after it has
convolved with a Gaussian kemel of the same size
but having a standard deviation of 3. As expected
the image is more smoothed
Figure (2): Smoothed Image 1
than that in Figure 2. The houses now appear as
blobs with their actual shapes becoming
unrecognizable. Smoothing is advantageous
because non-building objects smaller than
buildings are being
removed from the scene - simplifying the
recognition task.
The advantages of image scaling can be
practically demonstrated by applying the
homogeneous operator [7] to the source image
(Figure 1), as well as to a smoothed version
(Figure 2). The homogeneous operator detects
regions of
uniform intensity in the image. This is useful if the
objects being extracted have a uniform
appearance.
In both figures the roads, house roof panels and
barren sand patches are all identified by the
operator because of their uniform appearance.
However, notice that the regions in Figure 5 are
far less fragmented than those in Figure 4.
The house roof panels have been detected in both
images However, in Figure 12 some of the roof
panels have begun to merge with surrounding
regions. This indicates that there is a limit to the
amount of scaling that can be performed in order
to improve object extraction, Overall, it is felt' that
the image scaling approach has much potential.
III. MORPHOLOGIAL OPERATORS
A. Theory
The method we propose to extract building
objects from VHR-images relies on the use of
binary mathematical morphology operators
which are based on set theory [13]. The two
Figure (3): Smoothed Image 2
Figure(3): Source Image Regions
NICE-2010
Acharya Institute of Technology, Bangalore-560090 28
operators in MM are the erosion (I S) and the
dilation (IS) respectively defined as:
( ) { }
( ) { }
2
'
2
:
:
I S x S I
I S x S I
=
=
with S' and Sx respectively denoting reflex ion and
translation by x of the set S. From these basic
operators it is possible to define more complex
operators as we will see throughout this paper. It
is composed of three main steps. The first one
consists in the input grey level image binarization.
. The second step is an automatic morphological
filtering intended
Figure (4): Smoothed Image Regions
to eliminate some objects in the image and to
determine the size of the structuring elements.
The third step is the building extraction step itself
based on the use of an adaptive Hit-or-Miss
transform. Each of these steps is
Fig(5).VHR Images containing buildings
B. Generation of binary images
To generate binary image B, histogram of input
image is plot and from histogram threshold value
T related to min value between two maxima is
found out .:
( ) ( ) 1 if
0 otherwise
B x x T =
(0.3)
C. Morphological filtering
Before to extract the buildings from the generated
images, an automatic morphological filtering is
also performed. The aim of this filtering is to
remove objects whose size is lower than the
minimum size
Fig.2. Output of hit and miss image
of a building in the raw image. These objects may
be seen as noisy data capable of disturbing the
extraction process. The filtering used is a
morphological opening
defined as a combination of erosion and
dilatation:
( ) ( )
s
I I S S = (0.4)
D. Building detection
For building detection we propose to use the Hit
Or Miss Transform (HMT) which consists in a
double erosion of the image I and its complement
I
C
(i.e. the background) with two disjoint
structuring elements E and F. This transform is
particularly useful for template matching and is
defined as:
{ } ) ) (( ) ) (( :
) ( ) ( ) , )( (
C
X x
C
I F I E x
F I E I F E HMT I
=
=
(0.5)
Since we try to detect square or rectangular
buildings of various sizes, we adapt the HMT to be
able to take into account some structuring
elements E and F with varying sizes and shapes.
Our adaptive HMT is defined as:
U
L
K
C
F I E I F E HMT I ) ( ) ( ) , )( (
, ,
=
(1.6)
NICE-2010
Acharya Institute of Technology, Bangalore-560090 29
Thus, the result of this HMT is defined as the
union of all the results of the transform applied
with a given pair of structuring elements. The two
variable structuring elements Ea,b and Fc,d are
respectively defined as a rectangle of size a x b
and a frame of size c x d, with the constraints c >a
and
Fig.3. Output of extracted image
d > b. The sets and contain respectively all the
possible heights and widths of the SE, and a is a
coefficient used to determine the uncertain
area between E and F. In other words, it helps to
mark the area between pixels which surely belong
to buildings and pixels which surely belong to
background At the end of this operation, if the
parameters of the HMT have been correctly
defined, only the buildings are retained with their
respective position. However, the shape of these
buildings does not correspond any more to the
initial shape. Indeed, the HMT is based on
erosions which reduce the size of the objects.
Thus, a post processing is necessary to rebuild the
shape of the detected buildings. An additional
morphological operator is used for this task which
corresponds to a reconstruction .Using two
images, an input image I and a marker image M,
and applying until convergence a conditional
dilation with SE B defined as:
( ) T B M B M
T
=
(1.7)
Figure (2) shows the output of hit and miss image
in which some information is miss. Figure(3)
shows the final extracted output image which
gives the more information which is lost in hit and
miss transform.
IV CONCLUSION
The feasibility of applying scaling and
morphological operator for the extraction of
buildings from monocular grayscale aerial
photography has been examined.
Scaling an image reduces its information content.
Both noise and meaningful information are
removed due to scale-space events. The
suppression of meaningful information has to be
shown to actually improve object detection in
some instances. This is because redundant
substructure is removed, emphasizing the objects
themselves. The abstraction achieved by scaling
offers the potential to develop object model:; that
more closely match the actual objects. This is
because models based on abstracted objects can
be less rigid - there is no need to model object
substructure and object details. For instance, if an
image is smoothed enough a square house and
round hut will appear almost identical. A single
model can be developed to find the blobs - specific
object shape has ceased to be criterion.
We extended our previous works [12] by
introducing a bidimensional anemometry in the
filtering step. This morphological profile helps to
define automatically the structuring elements
used in the hit or miss transform. The
morphological operator produces a far better
Segmentation than the homogeneous operator in
certain classes of aerial photographs.
REFERENCES
[1] Hodgson, M.E., Window Size and and Visual
Image Classification Accuracy: An
Experimental Approach, ASPRS/ACSM
Technical Paper, 1994.
[2] Bums, J.B., Hanson, A.R. and Riseman, E.M.,
Extracting Straight Lines, IEEE Transaction
on Pattern Anafysisand Machin Intelligence,
vol. PAMI- 8, no. 4, July, 1986,pp 425-455.
[3] Venkatesww, V. and Chellapa, R., Extraction
of Straight Lines in Aerial Images, IEEE
Transactions on Partern Analysis and Machine
Intelligence, vol. 14, no. 11, Nov, 1992, pp 11
11-1 114.
[4] Huertas, A. and Nevatia, R., Detecting
Buildings in Aerial Images, Computer Vision,
Graphics and Image Processing, vol. 41, 1988,
pp 131-152.
[5] Mohan, R. and Nevatia, R., Using Perceptual
Organization to Extract 3- D Structures,
NICE-2010
Acharya Institute of Technology, Bangalore-560090 30
IEEE Transactions on Pattern Ana&srs and
Machine Intelligence, vol. 11, no. 11, Nov,
1989, pp 1121 1139
[6] Mayer. H., Abstractlion and Scale-Space
Events in Image understand, International
Archives of Photogrammetry and Remote
Sensing, vol 31, part 3, Vienna, 1996, pp 523-
528.
[7] Levitt, S. and Aghdasi, F., Texture Measures
for Building Recognition in Aerial
Photographs, Proceedings of COMSIG 1997,
Grahamstown South Africa, Sept, 1997, pp 75-
80.
[8] S. Lhomme, C. Weber, D. He, and D. Morin,
"Building extraction from vhrs images," in
ISPRS Congress, Istanbul, Turkey, 2004.
[9] I. Destival, "Mathematical morphology applied
to remote sensing," Acta Astronautica, vol. 13,
no. 6/7, pp. 371-385, 1986.
[10] E Laporterie, G. Flouzat, and 0. Amram,
"Mathematical morphology
[11] multi-level analysis of trees patterns in
savannas," in IEEE International
[12] Geosciences And Remote Sensing
Symposium, 2001, pp. 1496-1498.
[13] P. Soille and M. Pesaresi, Advances in
mathematical morphology applied to
geoscience and remote sensing," IEEE
Transactions on Geoscience and Remote
Sensing, vol. 40, no. 9, pp. 2042-2055,
September 2002.
[14] J. Weber, S. Lefevre, and D. Sheeren,
"Building extraction in vhrs images with
mathematical morphology," in International
Conference on Spatial Analysis and Geomatics,
Strasbourg, France, September 2006.
[15] J. Serra, Image Analysis and Mathematical
Morphology. Academic Press, 1982.
NICE-2010
Acharya Institute of Technology, Bangalore-560090 31
AUTOMATIC DETECTION OF HUMAN SKIN ALLERGYS USING IMAGE
PROCESSING TECHNIQUE
Mr. Shamshekhar Patil., Asst prof., Dept of CSE, Dr. Ambedkar Institute of Technology, Bangalore
Mrs. Shiva leela S.C., Lecture, Dept of MCA, Dr. Ambedkar Institute of Technology, Bangalore
Mr. Arvind C.S., M.Tech, Dr. Ambedkar Institute of Technology, Bangalore
csarvind2000@gmail.com
ABSTRACT
This paper is to develop a system that will
automatically perform and evaluate common skin
allergy tests on the human arm. The system has
two main branches of development; (1) The image
processing and vision system for guidance and
result evaluation and (2) The expert system for
classification of results. This paper is a progress
report for the vision system and the image
processing algorithms. It presents a method for
preliminary image processing for image
enhancement followed by the main digital
processing section which includes items such as
corrections for non-uniform illumination, hair
removal, adaptive thresholding and morphological
issues.
Index Terms: correlation method, adaptive
thresholding, allergy.
1. INTRODUCTION
The design of image processing and vision
system branch of the diagnosis system is not only
demanding and crucial for the performance of the
system as a whole, but also critical by medical
terms, because the test must be performed on
specific skin areas of the arm that are designated
with the help of anatomic criteria. For example,
the avoidance of veins is critical because a strong
reaction to a stimulant may cause a potentially
lethal allergic shock to the patient. Therefore,
careful image processing and test planning is
essential for the design of the whole system.
The first step in determining candidate
locations for the allergy agent placement is to
scan the skin and exclude the areas that have
lesions, wounds and veins. Previous research [1]-
[7] has shown the effectiveness of vein imaging
via the use of infrared (IR) illumination. The same
approach has been used in our system, but the IR
image is supplemented with normal visual images
and augmenting visible, infrared and laser
illumination.
For the complete automation of the
allergy test, the help of a machine vision system is
a strong prerequisite. The machine vision system
needs to address the following issues: i) Detection
of areas not suitable for performing the test. ii)
Monitoring of the subject position and location for
the safe and accurate guidance of the arm. iii)
Selection of the areas for stimulant dispensing. iv)
Evaluation of reactions with respect to the blood
concentration (erythema). v) Evaluation of
reactions in case of rash development.
2. IMAGING SYSTEM
The imaging system will consist of two
cameras, one for the IR imaging and one for the
normal light imaging. The IR camera is used
primarily at the first stage of the test for the
determination of vein location and any unusual
concentration of blood on the skin. The secondary
use of the infrared camera is for the detection of
the reaction results after the placement of the
allergy reactors. Any concentration of blood that
will show up as an erythema, will also be visible in
IR as a dark blob, because of the absorption of IR
light from the blood hemospherin. The visible
light camera serves multiple purposes: For
measuring the distance of the arm to the camera
via calibrated laser beam marks projected on the
object; For measurement of real-life dimensions
of the reactions on the skin; And for scanning
three dimensional anaglyph of reactions in cases
of blisters appearing via help of a projected laser
line on the skin.
The illumination of the work area is
performed via both IR and visible (white) light
LED sources that are sufficiently diffused via
air/polymer diffusers. Reflections are minimised
by polarizing filters in the sources and cameras,
oriented in perpendicular directions. The Infrared
optical system setup is shown in Fig2.1. The
sensitivity of the IR camera is selectively limited
to the IR region via proper rejection filters that
will not allow visible light in the camera. In
Fig.2.2, sample image from a working prototype
setup is shown, clearly displaying the benefits of
IR imaging for subcutaneous vein detection.
_____________________________________________________________________________________________________________________________
NICE-2010
Acharya Institute of Technology, Bangalore-560090 32
1 *
*
)) , ( (
) , (
*
)) , ( (
1 1
1 1
1 1
(
(
(
(
(
(
(
(
(
(
=
= =
= =
= =
N M
N M
y x I
y x I
N M
y x I
Tb
M
x
N
y
M
x
N
y
M
x
N
y
Figure 2.1: Infrared optical system.
Figure 2.2: Capture with IR source
illumination.
3. PROPOSED SYSTEM
3.1 Camera Calibration
In order to be able to perform measurements
on the allergy reactions, the cameras have to be
calibrated so that all real-lens projection
distortion on the image can be compensated in
the image processing software. Camera
calibration is the process that it is used to
estimate the intrinsic and extrinsic parameters of
the camera. In most cases the reliability and the
output of a machine vision system depends on the
accurate definition of the above parameters. The
calibration can be performed once and after the
camera parameters are estimated they can be
reused in the calculations, since the optical system
remains invariant. The estimation of the intrinsic
camera parameters is achieved by presenting
printed checkerboard images of known
dimensions to the camera. From these known
images the camera model can be iterated until the
parameters are estimated to sufficient precision.
Extrinsic camera parameters can be estimated by
varying the distance and angle of the presented
patterns to the camera. After the camera
calibration is performed, images acquired with
the cameras can be compensated for the lens
distortions and the measured image distances can
be correlated to the real world distances on the
patient skin.
3.2. Image Acquisition
At the current stage of the project the
image processing algorithms are prototyped and
developed in MATLAB. We have assumed that
the arm can be held still for about one second, Of
course, the capture rate depends on camera
characteristics and specifications. The multiple
images are then averaged as a first step for noise
removal. Averaging a number of images creates a
clearer picture, with less noisy pixel values [8].
3.3 Background Detection
For the optimal extraction of the test
points, the calculations need to be constrained to
the actual image area that contains the subjects
skin. Therefore, a special setup with a matte black
backdrop is used to ease the exclusion of the
background pixels from the calculations. The
background pixels can therefore be extracted with
the use of simple global thresholding, and the
threshold Tb that was used was set at the level of
the mean value of the image minus the standard
deviation, Equation (1), i.e for the initial greyscale
image I(x,y) ,where (x,y) are the spatial
coordinates, with dimensions MxN pixels, the
threshold Tb equals:
Eq. 1
Afterwards, the greyscale image is
converted to binary (bw) using the calculated
threshold Tb
>
=
Tb j i I if
Tb j i I if
j i bw
) , ( , 1
) , ( , 0
) , (
Eq.2
where (i, j) are the coordinates of each
image pixel. The result of the above process is
shown in Fig.3.1. Knowing the background pixels
coordinates, the algorithm processes only the
forearms pixels that reduces the execution time
and improves the veins detection procedure.
NICE-2010
Acharya Institute of Technology, Bangalore-560090 33
Figure 3.1: Background Detection
3.3 Hair Removal
Before we can proceed with the vein
detection we must first remove all the elements
that may interfere with the final result. Such
elements are the hair that may be present. We
remove the traces of the hair via morphological
closing. The result of this process is shown in
Fig.3.2. The selection of the proper structural
element (SE) to be used in opening the image is
very important [9]-[10]. Experiments have shown
that a vertical and a horizontal line SE in the
shape of cross should be used because of the
irregular orientation of the hair. The size of the SE
is estimated on a per-image basis from the
average hair thickness measured on the images.
Figure 3.2: Hair Removal.
3.3 Uneven Illumination Correction
One of the most frequent problems in
image processing is uneven illumination. The
result of the undesirable non-uniform
illumination is regions of the image being dark
and their contrast too low. The human eye has no
trouble perceiving the images, because the brain
will compensate the differences, taking into
account the ambient illumination. Unfortunately,
the computer cannot automatically compensate
for this effect, so the next goal of the processing is
to flatten the illumination component in the
image. An image is a function consisting of two
coefficients the reflectance and the illumination
which are combined in a non-linear manner [11].
The desired goal is to remove the illumination
gradient from the image and leave only the
objects reflectance.
The illumination component can be found
at the low frequencies of the image. In order to
calculate this component the image is filtered
with a very low pass filter in frequency domain.
This way, an estimation of the background of the
image is created, which is dominated by the
illumination coefficient. To eliminate any border
effects, the images are transformed from- and to-
the frequency domain via symmetric extension of
the image boundaries. The result of the low pass
filtering is the original image subjected to heavy
smoothing. Finally, the reflectance component of
the image can be calculated by dividing the
original (Fig.3.2) image with the smoothed image.
The division is the only way to extract the
reflectance coefficient because it is combined with
the illumination signal by multiplication (Fig.3.3).
Figure 3.3: Un even illumination correction.
3.4 Contrast Enchantement
The image produced after the illumination
correction has very low contrast. In order to
enhance the veins, an intensity transformation
was used which is called contrast stretching. This
transformation function compresses the input
levels lower than m in to a narrow range of dark
levels in the output image; similarly, it
compresses the values above m in to a narrow
band of light levels in the output. The value of m
could be calculated experimentally to optimize
the results. In this case the m coefficient is
calculated using statistical elements of the image
I(x,y).
1 *
*
)) , ( (
) , (
*
*
)) , ( (
1 1
1 1
1 1
(
(
(
(
(
(
(
(
(
(
+
(
=
= =
= =
= =
N M
N M
y x I
y x I
b
N M
y x I
m
M
x
N
y
M
x
N
y
M
x
N
y
Eq.3
NICE-2010
Acharya Institute of Technology, Bangalore-560090 34
The contrast stretching transformation is
described by the limiting function
E
r m
r T
) / ( 1
1
) (
+
=
Eq. 4
where r represents the intensities of the input
image, T(r) the corresponding intensity values
in the output images and E controls the slope of
the function. Fig.3.4 depicts the contrast
stretching effect. It must be noted that the
algorithm does not take into account the
background regions that were identified at the
background detection process [11].
Figure 3.4: Contrast Enhancement.
3.4 Veins Extraction
After the contrast enhancement
procedure, the pixels that they belong to the veins
have dark intensity levels than the pixels that they
belong to the rest skin area of the forearm.
Although that enhancement conduces to the
extraction of the skin areas that are suitable for
performing the allergy test, some problems of the
images segmentation may be still remained. This
situation is caused because of the non-uniform
intensity levels of the pixels that shape the veins.
So the use of a global thresholding method will
not produce satisfactory results. Many methods of
global, adaptive and dynamic thresholding
calculation have been tested in order to extract
the optimal shape of the veins. Finally, a
multilevel thresholding method is used by
producing satisfactory results.
Firstly, the profile of the pixels intensity
levels of each row of the image is computed. After
intensity profiling the maximum and the
minimum value of each distribution is calculated,
Fig.8. These values are the transition limits of the
veins pixels values. The maximum and minimum
values from the previous stage of the process are
averaging to estimate the down and the upper
limiting levels of the multilevel thresholding,
equation (5) and (6).
1=mean(min)-b*(mean(max)-mean(min)) 5
2=mean(max)-b*(mean(max)-mean(min)) 6
Afterwards, the greyscale image is
converted to binary (bw) using the
calculated threshold levels
=
else
j i bw
, 0
2 ) y I(x, 1 , 1
) , (
Eq. 7
where (i, j) are the coordinates of each
image pixel. The result of the above process
is shown in Fig.3.5. The value coefficient b
was determined by experiment and set to a
value of 0.38 for best results with the data
set used.
Figure 3.5: Veins extraction
3.5 Test point dispersion
The most crucial stage of the application
is the selection of points suitable for performing
the allergy test. These points must satisfy rules
that are set according to the standard medical
procedure for allergic tests. Let d denote the
maximum diameter that a test result can reach
(we assume that the test results will be circularly
shaped). The minimal distance of two adjacent
test points must be greater than d to avoid
overlapping of the test results. Another factor to
consider in the algorithm is that certain areas in
the hand forearm such as veins, gland, scratches
etc, must be avoided. In order to build a fast
dispersion algorithm and to ensure that the test
points will not overlap undesirable areas, a
morphological operation of dilation is used. A
structuring element with a disk shape is applied
because of the roughly circular shape of the
allergic test results. Its size must be shortly larger
than the estimated maximum diameter d of an
allergy reaction to secure a safe distance from the
prohibited areas. The examination of the test
points is accomplished through the distance
calculation between any two possible points. For
every new candidate test point the distance
between it and all of the previous points that have
already been set and registered is calculated. The
result of this process is shown in figure 3.6
NICE-2010
Acharya Institute of Technology, Bangalore-560090 35
Figure 3.6: Test point dispersion.
4. READING SKIN TEST REACTIONS
Following the allergens reaction on the skin is the
stage of results recognition. Ten to twenty
minutes after the placement of the allergens at the
designated positions of the skin, the results of the
reactions appear. An allergen reagent causes a
small blister that is called a wheal, and a red
region, which is called an erythema, appears
surrounding the wheal. To obtain best results the
dimensions of the allergen reactions are
measured under high illumination.
The mean values of the maximum vertical and
horizontal diameter of the wheal and erythema
are calculated. These values are used to grade the
reactivity of the allergic response. The grading is
implemented by scoring system [12]. The exact
shape of the wheal can be recovered in 3D via the
scanning of the area with a projected laser line.
This helps pinpoint the exact boundary of the
wheal without being distracted by any color
variations of the skin or the erythema.
5. CONCLUSION
In this paper we have presented a computer aided
testing and diagnosis system for allergy testing.
We have focused mainly on the design of the
vision and image processing system which is
based on the principle that the bloods
hemospherin absorbs IR lighting, therefore
making the detection of regions with high-level
blood concentrations, like veins, easier. The
development of this system is still in its early
stages and the work is still in progress; therefore,
some described features may change for the
benefit of speed and accuracy. In the current
stage of development, several methods are being
evaluated on a laboratory prototype setup to
determine the optimal procedure for allergy
detection (results) and classification.
6. REFERENCES
[1] H. D. Zeman, G. Lovhoiden and H. Desmhukh,
Optimization of subcutaneous vein contrast
enhancement, Proc. SPIE 2000 Biomedical
Diagnostic, Guidance and Surgical-Assist Systems
II, vol.3911, pp.50-57, May 2000.
[2] H. D. Zeman, G. Lovhoiden and H. Desmhukh,
Design of a Clinical Vein Contrast Enhancing
Projector, Proc. SPIE 2001 Biomedical
Diagnostic, Guidance and Surgical-Assist Systems
III, vol.4254, pp.204-215, June 2001.
[3] G. Lovhoiden, H. Desmhukh and H. D. Zeman,
Clinical Evaluation of Vein Contrast
Enhancement, Proc. SPIE 2002 Biomedical
Diagnostic, Guidance and Surgical-Assist Systems
IV, vol.4615, pp.61-70, May 2002.
[4] G. Lovhoiden, H. Desmhukh, C. Vrancken, Y.
Zhang, H. D. Zeman and D. Weinberg,
Commercialization of Vein Contrast
Enhancement, Proc. of SPIE 2003 Advanced
Biomedical and Clinical Diagnostic Systems,
vol.4958, pp.189-200, July 2003.
[5] G. Lovhoiden, H. Desmhukh and H. D. Zeman,
Prototype vein contrast enhancer, Proc. of SPIE
2004 Advanced Biomedical and Clinical
Diagnostic Systems II, vol.5318, pp.39-49, July
2004.
[6] Ph. Schmid and S. Fischer, Colour Segmentation
for the Analysis of Pigmented Skin Lesions,
Proc. of the Sixth International Conference on
Image Processing and its Applications 1997, vol.2,
pp.688-692, July 1997.
[7] Ph. Schmid-Saugeon, J. Guillod and J. P. Thiran,
Towards a computer-aided diagnosis system
for pigment skin lesions, in Computerized
Medical Imaging and Graphics, vol.27, No.1,
pp.65-78, 2003.
[8] R. C. Gonzalez, R. E. Woods, Digital Image
Processing, Prentice Hall, New Jersey, United
States of America, pp.10-18, 2002.
[9] T. Lee, V. NG, R. Gallagher, A. Coldman,
D.McLean, Dullrazor
: A software approach to
hair removal from images, Computers in Biology
and Medicine, vol.27, no 6, pp. 533-543,
November 1997.
[10] P. Soile, Morphological Image Analysis:
Principles and Applications, Springer, Berlin,
Germany, pp. 105-133, 2003.
[11] R. C. Gonzalez, R. E. Woods, Digital Image
Processing, Prentice Hall, New Jersey, United
States of America, pp.28-31, 2002.
[12] R. G. Slavin, R. E. Reisman, Expert Guide to
Allergy & Immunology, American College of
Physicians, pp.44, 1999.
NICE-2010
Acharya Institute of Technology, Bangalore-560090 36
ENERGY-EFFICIENT BIOTELEMETRY SYSTEM WITH NANO IP
Mr.R.Mohan,M.E
1
., P.Baranidaran
2
, K.Sudha
3
Sr. Lecturer/ECE
1
, Final year B.E-ECE
23
M.P.N.M.J.Engineering College, Chennimalai, Erode.
baranidaranp@yahoo.in
2
, sudha88@gmail.com
3
_____________________________________________________________________________________________________________________________
ABSTRACT
Technical advancements in embedded
systems, wireless communications and
physiological sensing allow small size, light weight,
ultra low power, and intelligent monitoring devices.
A number of these devices can be integrated into
energy efficient biotelemetry system with Nano IP,
a new enabling technology for health monitoring,
sports and military applications. The sensing unit
which is wearable device consists of sensors and
integrated circuits. These devices are capable of
performing health monitoring activities such as
heart rate, breath rate, blood glucose, body
temperature and nerve stimulus with a personal
computer and then monitored parameters can be
send to necessary place through Nano IP
communication protocol suit. An energy-efficient
implementation of Wireless Body Sensor Networks
(WBSNs) with embedded technology is designed
featured with Work-on-Demand protocol.
Dedicated for ultra-low-power wireless sensor
nodes, the system consist of a low power
Microcontroller unit (MCU), a Power Management
Unit (PMU), reconfigurable sensor interfaces and
communication ports controlling a wireless
transceiver with GSM technology. The MCU,
together with the PMU, provides quite flexible
communication and power-control modes for
energy efficient operations. This system will
consumes less amount of power than its existing
models, which is less than 3.3 volts. The measured
parameters are interfaced with mobile devices
through nano IP technology
1. INTRODUCTION
The greatest problem faced by most
wireless sensor networks is energy. When a
sensor is depleted of energy, it can no longer fulfill
its role unless the source of energy is replenished.
Therefore, it is generally accepted that the use of a
wireless sensors expires when its battery run out.
Here, we provide a solution to extend the
battery lifetime and to reduce the power
consumption through a technology named energy
harvesting. This can be achieved from various
sources such as vibrational energy, kinetic energy
that is produced from the body movements, body
temperature and solar energy.
One method of harvesting vibrational
energy is through the use of piezoelectric crystal.
Here, we are going to implement this efficient
energy harvesting technology in our proposed
system. This energy harvesting is used to reduce
the power consumption and to reduce the cost of
implementation.
2. NANO IP
In our system, the communication will be
made through the Nano IP. Nano IP stands for the
Nano Internet Protocol which is a concept that
was created to bring internet- like networking
services to embedded and sensor devices, without
the overhead of TCP/IP. Nano IP was designed
with minimal overheads, wireless networking and
local addressing in mind.
Fig 1: Nano IP architecture
The protocol actually consists of two transport
techniques, nanoUDP, which is an unreliable
simple transport, and nanoTCP, which provides
retransmissions and flow control. A socket-
compatible API is provided which makes the use
of the protocols very similar to that of IP
protocols. The only difference is in addressing and
the port range. NanoIP makes use of the MAC
address of underlying network technology rather
NICE-2010
Acharya Institute of Technology, Bangalore-560090 37
than IP addresses, which are not needed for local
networks. The port range is 8-bits, 256 ports each
for source and destination.
With the Nano Socket family, only a few
hours are required to add full-featured Internet
access functionality over LAN or WiFi to an
embedded device, including TCP sockets, SSL
encryption, routing, e-mail, and file transfers. The
logical interface between the host application and
the modules is AT+i Protocol, a simple text-
based API that enables fast and easy
implementation of Internet networking and
security protocols. The Nano Socket family
includes a plethora of security features. The
modules serve as a communications offload
engine and inherent firewall, protecting the
embedded device from the Internet attacks.
This module provides high level of
Internet security and encryption algorithms (AES-
128/256, SHA-128/192/256, 3DES, SSL3/TLS1
protocol for a secure client socket session) for
complete, end-to-end encryption support. The
Nano Socket iWiFi also includes the latest WLAN
encryption schemes (64/128-bit WEP,
WPA/WPA2 enterprise).
Nano Socket offers much more than many
other device servers on the market. It acts as a
security gap between the host application and the
network in Full Internet Controller mode. It
supports upto 10 simultaneous TCP/UDP sockets,
two listening sockets, a web server with two
websites, SMTP and POP3 clients, MIME
attachments, FTP and TELNET clients, and
SerialNET mode for serial to IP bridging. It
supports multiple Certificate Authorities and both
client-side and server-side authentication.
Nano Socket reduces the need to redesign
the host device hardware. Minimal or no software
configuration is needed for Nano Socket to access
the Wireless LAN. The serial operating mode
offers a true plug and play mode that eliminates
any changes to the host application. The Nano
Socket iWiFi also includes the latest WLAN
encryption schemes (64/128-bit WEP,
WPA/WPA2 enterprise).
The Nano Socket family eases
development by including USB, SPI, USART and
RMII interfaces. Both modules share the same
simple header-based pin out, allowing reduced
assembly costs and increased flexibility when
designing solutions. A single product design can
easily accommodate both LAN and WiFi.
The modules support several modes of
operation:
LAN-to-WiFi Bridge allowing transparent
bridging of LAN over WiFi, using direct RMII
connection to existing MAC hardware or PHY-
to-PHY connection.
Serial-to-LAN/WiFi Bridge allowing
transparent bridging of serial data over LAN
or WiFi.
Full Internet Controller mode allowing
simple microcontrollers to perform complex
Internet operations such as e-mail, FTP, SSL,
and others.
PPP emulation allowing existing (e.g.,
cellular modem) designs currently using PPP
to interface to the cellular modem by
connecting transparently over LAN or WiFi
with no changes to application or drivers.
Following are the additional features of
Nano Socket:
Non-volatile, on-chip operational parameter
database
Supports infrastructure and ad-hoc Wireless
LAN networks
SerialNET mode for Serial-to-IP bridging(port
server mode)
Local Firmware update
Remote configuration and firmware update
over the Internet
Retrival of time data from a Network Time
Server
The typical applications of Nano IP Socket
are adding IP communications over WiFi or LAN
to serial embedded devices and adding SSL
security to M2M solutions. Nano socket firmware
is remotely updateable and thus new security or
connectivity protocols do not require application
redesign, increased memory, or faster processor
speeds.
3. ENERGY HARVESTING
A wireless sensor network that is not
dependent on a limited power source (like a
battery) essentially has infinite lifetime. Much of
the research on wireless sensor networks has
assumed the use of a portable and limited energy
source, namely batteries, to power sensors and
focused on extending the lifetime of the network
NICE-2010
Acharya Institute of Technology, Bangalore-560090 38
by minimizing energy usage. Failure due to other
causes (like structural hardware damage) can be
overcome by self-organization and network re-
configuration. This has motivated the search for
an alternative source of energy to power wireless
sensor networks especially for applications that
require sensors to be installed for long durations
(up to decades) or embedded in structures where
battery replacement is impractical.
The harnessing and collection of ambient
energy into a useful form is called Energy
Harvesting. The use of ambient energy to generate
electricity is not a new concept. The ambient
energy that is being harvested to generate
electricity nowadays includes solar, wind,
mechanical and thermal energy. Harvesting
energy for low power devices like wireless
sensors parents is a big challenge as the energy
harvesting device has to be compatible in size (i.e.
small enough) with the sensors. There are
complex tradeoffs to be considered when
designing energy harvesting circuits for wireless
sensor networks arising from the interaction of
various factors.
Following are the some benefits of Energy
Harvesting:
Reduce dependency on battery power as
harvested energy may be sufficient to
eliminate battery completely.
Reduce maintenance costs as Energy
harvesting allows for devices to function
unattended and eliminates service visits to
replace batteries.
Providing sensing and actuation capabilities
in hard-to access hazardous environments on
a continuous basis.
Reduce environmental impact of hazardous
chemicals.
The more efficient and safety energy
harvesting technology is the energy harvesting
from vibration energy through the use of
piezoelectric capacitor. This method is the safest
and efficient method in the sense, this power
generated from vibration is not harmful as even
solar energy is not safe at all instances to human
body. In a vibration based harvesting, micro
power generator is used to scavenge body
vibrations for use in the sensor node.
Experimental results have shown that when a
piezoelectric crystal is depressed, sufficient
energy is harvested to transmit two complete 12-
bit digital words wirelessly.
Similarly, a system that harvests energy
from the forces exerted on a shoe during walking
has been demonstrated and indoor locations, like
staircases, are potential locations to harvest
vibrational energy for powering body sensors.
4. SYSTEM ARCHITECTURE
Fig 2: Functional block diagram of energy efficient
biotelemetry system.
The Bio-Telemetry system consists of body
sensor networks in which slave sensor nodes can
be used for bio-medical information acquisition;
signal preprocessing, data storage, and wireless
transmission (sometimes direct transmission
without any preprocessing). These body sensors
are used to sense the various bio-medical
parameters such as breath rate, temperature of
the body, motion of body parts, glucose level in
the blood and also heart beat rate. The sensed
analog signals are converted to digital signals
appropriate for transmission. This type of slave
sensor node is called as the sensing node. In
addition, the function of sensor nodes can be
expanded to monitoring, diagnosing and for
treatment purpose and this type of slave sensor
node is called as the stimulating node. The sensed
signals are driven to master node through the
slave node. Then the master node transmits the
sensed signals to the desired computer system at
where monitoring of patient has been done. This
transmission can be performed through Nano IP
module.
Energy harvesting is achieved by placing
piezoelectric crystal to the moving body parts.
Piezoelectric crystal works on the principal of
piezoelectric effect, through which vibrational
energy produced by the body parts is converted to
electrical energy that gives sufficient additional
energy for the sensor unit. Energy harvesting
technology increases the life time of all the
sensors, and also reduces the power consumption.
NICE-2010
Acharya Institute of Technology, Bangalore-560090 39
The sensed signals are sent to the
microcontroller PIC 16F877A that transmits the
signals to the Nano IP. This Nano IP protocol acts
as a transceiver system and it helps to transmit to
the desired location and for receiving diagnose
signals from the doctor or physician, if the patient
is in abnormal condition. The abnormal condition
of the patient can be easily predicted with the
highly reliable patient monitoring through Nano
Internet Protocol Transmission. The buffer is used
to store the information temporarily and driver is
used to direct the necessary devices to perform
first-aid treatment by stimulating needles, syringe
dispenser, and by controlling valves and relays.
In addition, this system can perform the first-
aid treatment when the physiological parameters
value exceeds the safety level. For example, when
the breath rate of a patient goes down behind the
normal value, then the valve accompanied with
the oxygen cylinder will automatically supply the
air to the patient and the amount of oxygen that to
be supplied to the patient will be controlled by the
relay unit present in our system.
5. CIRCUIT IMPLEMENTATION
a. Power Management Unit:
The Power management Unit consists of
piezoelectric crystal, bridge rectifier and storage
battery. The piezoelectric crystal accompanied
with bridge rectifier to harvest energy. The
vibrational energy is converted here as AC voltage
by the crystal and then the bridge rectifier
produces the DC voltage. The harvested energy
supplied to the battery which can store energy
and also the battery receives energy from main
power supply. Thus the lifetime of the battery
extends.
Fig 3: Power Management Unit
b. Sensor Unit:
The body parameters which to be measured
are measured through respective sensors. The
heart rate is sensed by an indirect manner such as
using IR transmitter and IR receiver with a
reference voltage. The heart rate is nothing but
number of times the heart pumps the blood. The
receiver associated with microcontroller could
sense the blood flow per unit time. Similarly blood
glucose can be measured using IR LED and IR
LDR.
c. Stimulus Unit:
The stimulus unit is to perform first-aid
process whenever the physiological parameters
exceed its safety level.
Fig 4: Stimulus Unit
d. Communication Unit With Nano IP:
This protocol suit is mainly designed to
reduce the gap between networking parameters
and embedded like devices. The nano IP provides
highly reliable and highly securable network.
Performance specifications of nano IP:
Host data rate: up to 3Mbps in several mode,
12 Mbps in LAN WiFi mode.
Serial data format: asynchronous character;
binary; 8 data bits; no parity; 1,1.5,2 stop bits.
Serial NET mode : asynchronous character;
binary; 7 or 8 data bits; odd, even, or no
parity; 1, 1.5, or 2 stop bits.
Flow control: Hardware (-RTS,0CTS) and
software flow control.
Typical applications of nano IP:
Adding IP communication over WiFi to serial
embedded devices
Piezocrystal Bridge
Rectifier
Battery to all
units
Microcontroller
Unit
Buffer
Syringe
dispenser
Valve
Driver
NICE-2010
Acharya Institute of Technology, Bangalore-560090 40
Replacing a LAN cable with a WiFi connection
Adding SSL security to M2M solutions.
6. CONCLUSION
Hence implementation of the bio
telemetry system using Nano IP has been
proposed. Compared to the conventional
protocols used in bio-telemetry systems, the
proposed Nano Internet Protocol suit gives
efficient and reliable transmission of bio-medical
parameters to the monitoring location.
The power management unit in the
system consists of energy harvesting unit that
harvests energy using piezoelectric crystal, to
reduce the power consumption below 4V and also
to control the implementation expensive and
make it to available to the all categories of people.
7. REFERENCE
1. Fiocchi and C.Gatti, A very flexible BiCMOS
low-voltage high-performance source
follower, in Proc. ISCAS, 1999, vol.2, pp. 212-
215.
2. G.Z.Yang, Body sensor networks, London,
U.K:springer,2006.
3. J.F.Dickson, On-chip high-voltage generation
in MNOS integrated circuits using an
improved voltage multiplier technique, IEEE
Solid-State Circuits, vol. 11, no. 3,pp. 374-378,
Jun.1976.
4. R.Dudde and T.Vering, Advanced insulin
infusion using a control loop (ADICOL)
concept and realization of a control-loop
application for the automated delivery of
insulin, presented at the 4
th
Annu. IEEE Conf.
Information Technology Applications in
Biomedicine, Birmingham, U.K., 2003.
5. R.Puers and P.Wouters, Adaptive interface
circuits for flexible circuits for flexible
monitoring of temperature and movement,
Analog Integr. Circuits Signal Process., vol. 14,
pp. 193-206,1976.
6. Xiaoyu Zhang, Hanjun Jiang, Lingwei Zhang,
Chun Zhang, Zhihua Wang, and Xinkai Chen,
An energy efficient ASIC for wireless body
sensor networks in medical applications,
IEEE transactions on biomedical circuits and
systems, vol.4, No.1, February 2010.
NICE-2010
Acharya Institute of Technology, Bangalore-560090 41
BRAIN WAVE CONTROLLER FOR STRESS REMOVAL AND AUTOMATION OF
AUTOMOBILE IGNITION TO PREVENT DRIVING UNDER INFLUENCE
CONTROLLING BRAINWAVES USING EMBEDDED SYSTEMS
L. Sabarish
sabarish.lakshmanan@yahoo.com
Department Of Electronics and Electronics Engineering
Rajalakshmi Engineering College,
Rajalakshmi Nagar, Thandalam, Chennai 602 105, Ph : 9710350536
___________________________________________________________________________________________________________________________________
ABSTRACT
Stress is a prevalent and costly problem in
today's workplace. It is the harmful physical
and emotional response that occurs when there is a
poor match between job demands and the
capabilities, resources, or needs of the worker.
Persistence of stress results in cardiovascular
disease such as depression, concentration and
memory loss.
Addiction is one of the chronic disorders
that are characterized by the repeated use of
substances or behaviors despite clear evidence of
morbidity secondary to such use. It is a
combination of genetic, biological /
pharmacological and social factors. Example:
gambling, alcohol drinking, taking narcotic drugs
and certain mannerisms. The therapies at present
consume time.
About 24% of the accidents taking place
are due to drunken drive. A driver subjected to long
drive falls sleepy and ends up in accident.
In this paper we briefly discuss about the
brain wave and brains reaction during stress,
addiction and drunk. This paper also explains you
the basic task of Brainwave Controller, that how
stress, addiction is identified with the help of
brainwave and how these are controlled using the
principle binaural beats. Also we have designed a
device to detect the brainwaves and process it to
determine whether it is addiction or stress. In
addition to controlling of brainwaves, it also has a
feature to avoid an individual who consumes
alcohol to drive a vehicle. This paper promises to be
an economical solution for the people who suffer
from stress, addiction and to prevent accidents.
Keywords :
Stress, Brainwave, cardiovascular disease,
binaural beats, addiction, driving under influences
(DUI) and ignition interlock system or device
(IID).
I. INTRODUCTION
A. Brain:
Brain is an electro-chemical organ. The
Brainwaves are produced by the frontal lobe of
the brain. It processes auditory information from
the ears and relates it to Wernicke's area of the
parietal lobe and the motor cortex of the frontal
lobe. The amygdala is located within the temporal
lobe and controls social and sexual behavior
and other emotions. The limbic system is
important in emotional behavior and controlling
movements.
Fig.1 Side view of the human brain with
parts
Researchers have speculated that a fully
functional brain can generate as much as 10 watts
of electrical power. Even though this electrical
power is very limited, it does occur in a very
specific ways that are characteristic of the human
brain.
B. Brainwaves:
Electrical activity emanating from the
brain is displayed in the form of brainwaves.
There are four categories of these brainwaves,
Acharya Institute of Technology, Bangalore
ranging from most activity to least activity. These
are delta waves, theta waves, alpha waves and
beta waves. The waveform corresponding to these
categories are shown in figure 2.
1) Delta waves: These are waves with high
amplitude. It has a frequency of 0.5
Hertz. They never go down to zero
because that would mean that you were
brain dead. But, deep dreamless sleep
would take you down to the lowest
frequency. Typically, 2 to 3 Hertz.
2) Theta waves: These are waves with
amplitude lesser than that of delta wav
and have a greater frequency of 5
Hertz. A person who has taken time off
from a task and begins to daydream is
often in a theta brainwave state.
3) Alpha waves: These are waves with
amplitude lesser than that of theta waves
and have a greater freque
Hertz. A person who takes time out to
reflect or meditate is usually in a alpha
state.
4) Beta waves: These are the waves that have
the lowest amplitude and have the highest
frequency of 15 40 Hertz. These waves
are again classified into low
and high beta waves according to their
range of frequencies. The low beta waves
have a frequency of 15
person making an active conversation
would be in the low beta state. The
beta waves have a frequency of 33
Hertz. A person in a stress, pain or
addiction would be in the high beta state.
Figure 3 shows the representation of high
and low beta waves.
, Bangalore-560090
g from most activity to least activity. These
are delta waves, theta waves, alpha waves and
beta waves. The waveform corresponding to these
These are waves with high
amplitude. It has a frequency of 0.5 4
ertz. They never go down to zero
because that would mean that you were
brain dead. But, deep dreamless sleep
would take you down to the lowest
frequency. Typically, 2 to 3 Hertz.
These are waves with
amplitude lesser than that of delta waves
and have a greater frequency of 5 8
Hertz. A person who has taken time off
from a task and begins to daydream is
often in a theta brainwave state.
These are waves with
amplitude lesser than that of theta waves
and have a greater frequency of 9-14
Hertz. A person who takes time out to
reflect or meditate is usually in a alpha
These are the waves that have
the lowest amplitude and have the highest
40 Hertz. These waves
are again classified into low beta waves
and high beta waves according to their
low beta waves
32 Hertz. A
person making an active conversation
would be in the low beta state. The high
have a frequency of 33 40
rson in a stress, pain or
addiction would be in the high beta state.
Figure 3 shows the representation of high
S.No. Brainwaves
1) Delta
2) Theta
3) Alpha
4) Low Beta
5) High Beta
Table 1 Different brainwave and its frequencies
Fig. 2 Different brainwaves with their names and
the situations when it occurs.
Fig. 3 High beta waves and Low beta waves
respectively
C. Addiction:
There are two types of addiction:
dependency and Psychological dependency.
1. Physical dependency :
Physical dependence on a substance is
defined by appearance of characteristic
NICE-2010
560090 42
Frequency
range
(Hertz)
0.5 - 4
5 - 8
9 14
15 32
32 - 40
Table 1 Different brainwave and its frequencies
Fig. 2 Different brainwaves with their names and
the situations when it occurs.
Fig. 3 High beta waves and Low beta waves
respectively
There are two types of addiction: Physical
dependency and Psychological dependency.
Physical dependency :
Physical dependence on a substance is
defined by appearance of characteristic
NICE-2010
Acharya Institute of Technology, Bangalore-560090 43
withdrawal symptoms when the drug is suddenly
discontinued. Some drugs such as cortisone, beta
blockers etc are better known as
Antidepressants. Some drugs induce physical
dependence or physiological tolerance - but not
addiction - for example many laxatives, which are
not psychoactive; nasal decongestants, which can
cause rebound congestion if used for more than a
few days in a row; and some antidepressants,
most notably Effexor, Paxil and Zoloft, as they
have quite short half-lives, so stopping them
abruptly causes a more rapid change in the
neurotransmitter balance in the brain than many
other antidepressants. So a doctor should be
consulted before abruptly discontinuing them.
2. Psychological dependency:
Psychological addictions are a dependency
of the mind, and lead to psychological withdrawal
symptoms. Addictions can theoretically form for
any rewarding behavior, or as a habitual means to
avoid undesired activity, but typically they only
do so to a clinical level in individuals who have
emotional, social, or psychological dysfunctions,
taking the place of normal positive stimuli not
otherwise attained. Psychological addiction, as
opposed to physiological addiction, is a person's
need to use a drug or engage in a behavior despite
the harm caused out of desire for the effects it
produces, rather than to relieve withdrawal
symptoms. As the drug is indulged, it becomes
associated with the release of pleasure-inducing
endorphins, and a cycle is started that is similar to
physiological addiction. This cycle is often very
difficult to break.
We are going to solely consider the
psychological addictions in designing the
addiction avoider device.
D. Recovery Therapy from Addiction:
Some medical systems, including those of at least
15 states of the United States, refer to an
Addiction Severity Index to assess the severity of
problems related to substance use. The index
assesses problems in six areas: medical,
employment/support, alcohol and other drug use,
legal, family/social, and psychiatric. While
addiction or dependency is related to seemingly
uncontrollable urges, and has roots in genetic
predisposition, treatment of dependency is
conducted by a wide range of medical and allied
professionals. Early treatment of acute
withdrawal often includes medical detoxification,
which can include doses of anxiolytics or
narcotics to reduce symptoms of withdrawal. An
experimental drug, ibogaine, is also proposed to
treat withdrawal and craving. Alternatives to
medical detoxification include acupuncture
detoxification. In chronic opiate addiction, a
surrogate drug such as methadone is sometimes
offered as a form of opiate replacement therapy.
But treatment approaches universal focus on the
individual's ultimate choice to pursue an alternate
course of action. Anti-anxiety and anti-depressant
SSRI drugs such as Lexapro are also often
prescribed to help cut cravings, while addicts are
often encouraged by therapists to pursue
practices like yoga or exercise to decrease
reliance on the addictive substance or behavior as
the only way to feel good. Treatments usually
involve planning for specific ways to avoid the
addictive stimulus, and therapeutic interventions
intended to help a client learn healthier ways to
find satisfaction. Clinical leaders in recent years
have attempted to tailor intervention approaches
to specific influences that effect addictive
behavior, using therapeutic interviews in an effort
to discover factors that led a person to embrace
unhealthy, addictive sources of pleasure or relief
from pain.
E. Driving Under Influence(DUI):
Driving under the influence of alcohol (operating
under the influence, drinking and driving,
impaired driving) or other drugs, is the act of
operating a vehicle (including boat, airplane, or
tractor) after consuming alcohol or other drugs.
DUI or DWI are synonymous terms that represent
the criminal offense of operating (or in some
jurisdictions merely being in physical control of) a
motor vehicle while being under the influence of
alcohol or drugs or a combination of both.
It is a
criminal offense in most countries as it
contributes to some of the major accidents. Now
let us consider few basic concepts upon which this
project is largely based upon.
Blood alcohol content or blood alcohol
concentration (abbreviated BAC) is the
concentration of alcohol in a person's blood. BAC
is most commonly used as a metric of intoxication
for legal or medical purposes. It is usually
expressed in terms of volume of alcohol per
volume of blood in the body. That is a unit-less
ratio commonly expressed as parts per million
NICE-2010
Acharya Institute of Technology, Bangalore-560090 44
(PPM) or as a fractional percentage. That is a
decimal with 2-3 significant digits followed by a
percentage sign, which means 1/100 of the
previous number (E.g., 0.0008 expressed as a
percentage as 0.08%). Since measurement must
be accurate and inexpensive, several
measurement techniques are used as proxies to
approximate the true parts per million measure.
Some of the most common are listed here: (1)
Volume of alcohol per volume of exhaled breath
(E.g. 0.08 mL/L), (2) Mass per volume of blood in
the body (E.g.: 0.08 g/L), and (3) Mass of alcohol
per mass of the body (E.g.: 0.08 g/Kg). After one
drink you reach your peak after 30 minutes and
you should wait a few hours before you drive. The
number of drinks consumed is often a poor
measure of BAC, largely because of variations in
weight, sex, and body fat.
F. Ignition Interlock System:
An ignition interlock device or breath alcohol
ignition interlock device (IID and BIID) is a
mechanism, like a breathalyzer, installed to a
motor vehicle's dashboard. Before the vehicle's
motor can be started, the driver first must exhale
into the device, if the resultant breath-alcohol
concentration analyzed result is greater than the
programmed blood alcohol concentration, usually
0.02% or 0.04%, the device prevents the engine
from being started. At random times after the
engine has been started, the IID will require
another breath sample. The purpose of this is to
prevent a friend from breathing into the device,
enabling the intoxicated person to get behind the
wheel and drive away. If the breath sample isn't
provided, or the sample exceeds the ignition
interlock's preset blood alcohol level, the device
will log the event, warn the driver and then start
up an alarm (e.g., lights flashing, horn honking,
etc.) until the ignition is turned off, or a clean
breath sample has been provided. A common
misconception is that interlock devices will
simply turn off the engine if alcohol is detected;
this would, however, create an unsafe driving
situation and expose interlock manufacturers to
considerable liability. An interlock device cannot
turn off a running vehicle, all that an Interlock
device can do is interrupt the starter circuit and
prevent the engine from starting.
II. PRINCIPLE
The principle behind this device is
Binaural Beats. Binaural beats or binaural tones
are auditory processing artifacts, which are
apparent sounds, the perception of which arises
in the brain independent of physical stimuli. The
brain produces a similar phenomenon internally,
resulting in low-frequency pulsations in the
loudness of a perceived sound when two tones at
slightly different frequencies are presented
separately, one to each of a subject's ears, using
stereo headphones. A beating tone will be
perceived, as if the two tones mixed naturally, out
of the brain. The frequency of the tones must be
below about 1,000 to 1,500 hertz. The difference
between the two frequencies must be small
(below about 30 Hz) for the effect to occur;
otherwise the two tones will be distinguishable
and no beat will be perceived.
Binaural beats can influence functions of
the brain besides those related to hearing. This
phenomenon is called frequency following
response (FFR). The concept is that if one receives a
stimulus with a frequency in the range of brain
waves, the predominant brain wave frequency is
said to be likely to move towards the frequency of
the stimulus (a process called entrainment).
Human hearing is limited to the range of
frequencies from 20 Hz to 20,000 Hz, while the
frequencies of human brain waves are below
about 40 Hz. To account for this, binaural beat
frequencies must be used.
According to this view, when the
perceived beat frequency corresponds to any of
the brainwave frequencies, the brain entrains to
or moves towards the beat frequency. For
example, if a 315 Hz sine wave is played into the
right ear and a 325 Hz one into the left ear, the
brain is supposed to be entrained towards the
beat frequency 10 Hz. Alpha range is usually
associated with relaxation, this is supposed to
have a relaxing effect. Some people find pure sine
waves or pink noise unpleasant, so background
music (e.g. natural sounds such as river noises)
can also be mixed with them.
III.CONSTRUCTION AND WORKING
A. Block diagram
The General block diagram of controlling
addiction/stress is shown in the figure 4.
Acharya Institute of Technology, Bangalore
Fig. 4 General block diagram to control addiction /
stress
Fig. 5 General block diagram to avoid DUI
Fig. 6 General block diagram to avoid accidents
The block diagram used in implementation of
brainwave controller with all its modes is given in
figure 7.
, Bangalore-560090
Fig. 4 General block diagram to control addiction /
Fig. 5 General block diagram to avoid DUI
Fig. 6 General block diagram to avoid accidents
The block diagram used in implementation of
with all its modes is given in
Fig. 7 General block diagram of Brainwave
controller
B. EEG Sensors
EEG sensors is used to measure the
electrical equivalent signal of brain wave.It consist
of a 0.7 inch diameter hard plastic outer disc
housing with a pre-jelled Silver chloride snap style
post pellet insert. These sensors do not contain any
latex. Figure 8 shows the representation of Ag/Agcl
EEG sensor.
Fig. 8 Electroencephalography (EEG) sensors
Fig.9 EEG Fabricated Headset
1 Stress
2 Drunken drive
3 Night drive
R Reset
EEG
EEG
EEG
Amplifier
Amplifier
Amplifier
DSP Processor
Program
Program
Program
NICE-2010
560090 45
Fig. 7 General block diagram of Brainwave
controller
EEG sensors is used to measure the
electrical equivalent signal of brain wave.It consist
of a 0.7 inch diameter hard plastic outer disc
jelled Silver chloride snap style
post pellet insert. These sensors do not contain any
latex. Figure 8 shows the representation of Ag/Agcl
Fig. 8 Electroencephalography (EEG) sensors
Fig.9 EEG Fabricated Headset
Switch
1 2 3
Processor
Motor
Driving
DAC
Program
1
Program
2
Program
3
DAC
DC
Wake-
up
Right-
Side
Left-Side
Headpho
R
NICE-2010
Acharya Institute of Technology, Bangalore-560090 46
The sensor sends the analog brainwave signal into
the instrumentation amplifier circuit.
C. Instrumentation Amplifiers
The amplitude of analog brainwaves is in
between the range of 150 250 micro volts (V).
This is very low. For processing, at least
amplitude above 2 volt is needed. For this a high
gain and low noise amplifier is needed. For that
instrumentation amplifier with high gain and high
CMRR ratio is employed.
Here one operational is not enough to
produce this much high gain. So a series of
amplifier is cascaded to give required gain. The
gain of an individual inverting operational
amplifier is given by.
Gain (A) = -R2/R1
Fig.9 Circuit diagram of cascaded inverting
amplifier with a gain of 20,000. (Designed in
Pspice)
Here we are using four inverting amplifier
cascaded as shown in figure 9. Let the gain of each
inverting amplifier from left to right be A1, A2, A3
and A4. And let Vi and Vo be the input and output
voltages of the amplifier.
Now,
A1 = (-R2/R1)
= (-2/1)
= -2
A2 = (-R4/R3)
= (-10/1)
= -10
A3 = (-R6/R5)
= (-10/1)
= -10
A4 = (-R8/R7)
= (-100/1)
= -100
*all resistors are in kilo ohm
Now Total Gain of the amplifier (Aeff),
A eff = A1 * A2 * A3 * A4
A eff = (-2)*(-10)*(-10)*(-100)
A eff = 20, 000
Therefore,
Vo = Vi * Aeff * Vi
= 15 * 10
-5
* 20, 000 V
Vo = 3 Vi
Hence an amplifier with gain 20,000 is
designed using basic operational amplifier.
D. DSP Processor
The Amplified EEG signal is given to
TMS320C6713DSP Processor where it has a pre-
defined program to select a range of frequency
NICE-2010
Acharya Institute of Technology, Bangalore-560090 47
from 1 50 Hertz. Then that selected range is
converted from analog to digital samples. The DSP
processor has a definite program according to
different modes selected from the switch. There
are 3 modes with which we can use this device.
E. Simulation:
The following simulation images indicate
the recorded brainwave using 2 EEG electrodes:
1. Mode - 1: Controlling addiction / stress:
In mode 1 the DSP processor checks for
the frequency range between 32 40 Hertz. If the
range is between 32 40 Hertz, it is considered
that the person is under addiction/stress and the
processor runs a look-up table which contains the
digital samples of binaural beats. The samples
produce sine wave with a difference of 10 Hertz.
These two waves are sent to each side of
headphone.
Fig.10 Block Diagram Sending two similar tones
with difference in frequency.
This generates the binaural beats. This is
given to the ears . As the difference in these two
waves is 10 Hz which is below 20 Hz, it cannot be
detected by the human ear. But there is a neuron
called afferon neuron inside the ear which senses
this 10 Hz and sends it to brain as a stimulus. Now
this stimulus entrains the brain to generate a
stimulus of brainwaves similar to the supplied
stimulus thereby reducing the brainwaves from
32 40 Hz to 9 14 Hz making the mind relaxed.
2. Mode 2: Avoiding DUI :
When the switch is turned to mode 2, the
device is connected to internal circuitry of a
vehicle. Firstly, an Ignition Interlock Device (IID)
is placed in an automobile and it is made
mandatory for the driver to exhale in to the
breathalyzer to switch on the automobiles
engine. Here, the Blood Alcohol Level (BAC) of the
driver is analyzed and in case of high BAC the
engine does not start. If BAC is low, then a
specially fabricated Electroencephalograph (EEG)
headset (which contains EEG sensors according to
international 10 -20 system) should be placed in
the drivers head to analyze the drivers
brainwaves. By analyzing the drivers brainwaves,
the risk of driving under the influence of drugs is
reduced. Here the BAC is measured as per the
international threshold value of 0.04ml/L. Once
the BAC is declared low, then the driver has to
take up the drug test, which involves the usage of
EEG headset in order to detect the brainwave
activity of the driver. Unless and otherwise the
driver has his brain wave levels at alpha or beta
mode, the engine will not start.
3. Mode 3: Night drive:
NICE-2010
Acharya Institute of Technology, Bangalore-560090 48
When the switch is turned to mode 3, the
device is used to avoid sleep while long night
drive. Here the individual must wear a cap at all
times. This cap is embedded with EEG sensors as
explained in mode 2. Here too DSP processor
checks whether the signal has frequency below 7
Hertz which means the individual is nearly asleep.
If such is the case, the processor triggers an alarm,
so that accidents can be avoided. This alarm could
be of any form. For example it could be horns
honking or the audio system playing loud music
or enabling specific alarm device to perform the
waking up operation.
IV. CONCLUSION
We have designed a device that has three
purposes. Firstly, the brainwaves are controlled
by the principle of binaural beats and frequency
following response thereby controlling addiction
or stress by making the mind relaxed temporarily.
Secondly, the brainwaves are continuously
monitored to avoid drunken drive in a vehicle.
Thirdly, the brainwaves are continuously
monitored to falling asleep while driving long
distance in a vehicle. Though all the above
applications are discreet to each other, it is
absolutely useful to use a same device for all the
three purposes.
V. MERITS
1. The whole device is light weight and can
be carried anywhere.
2. The whole device including sensors and
headphone is cheap and costs only about
Rs. 1500 and slightly above.
VI. DEMERITS
Those meeting any of the following
criteria should not use binaural beats
Epileptics
1. Pregnant women
2. People susceptible to seizures
3. Pacemaker users
4. Photosensitive people.
VII. FUTURE ENHANCEMENTS
1. The concept of frequency following
response can be further researched to
ease communication with deaf and
dumb individuals.
2. The concept of binaural beats can be
further used to study the resonance of
brain during brain diseases.
VIII. REFERENCES
[1] Detection of seizures in epileptic and
non-epileptic patients using GPS and
Embedded Systems by Abhiram
Chakraborty Ukranian Journal of
Telemedicine and medical Telematics
(TOM 3 No.2 pp 211)
[2] Automated realtime interpretation of
brainwaves by Kenneth George Jordan
US patent
[3] Method and apparatus for changing brain
wave frequency by John L Carter US
patent 50368585036858
[4] en.wikipedia.org/wiki/ binaural beats
[5] en.wikipedia.org/wiki/Addiction
[6] en.wikipedia.org/wiki/Drug_Addiction
[7] www.trdrp.org/research/ PageGrant.asp?
grant_id=383
NICE-2010
Acharya Institute of Technology, Bangalore-560090 49
CLUSTER THE UNLABELED DATASETS USING EXTENDED DARK BLOCK
EXTRACTION
Srinivasulu Asadi
1
, Ooruchintala Obulesu
2
, P Sunilkumar Reddy
3
Dept of IT
12
, Dept of MCA
3
S.V.E.C, A.Rangampet, Tirupati - 517 502, India.
srinu_asadi@yahoo.com
1
, oobulesu681@gmail.com
2
, pg.sunilkumar@gmail.com
3
ABSTRACT
Clustering analysis is the problem of
partitioning a set of objects O = {o1 on} into c self-
similar subsets based on available data. In general,
clustering of unlabeled data poses three major
problems: 1) assessing cluster tendency, i.e., how
many clusters to seek? 2) Partitioning the data into
c meaningful groups, and 3) validating the c
clusters that are discovered. We address the first
problem, i.e., determining the number of clusters c
prior to clustering. Many clustering algorithms
require number of clusters as an input parameter,
so the quality of the clusters mainly depends on this
value. Most methods are post clustering measures
of cluster validity i.e., they attempt to choose the
best partition from a set of alternative partitions.
In contrast, tendency assessment attempts to
estimate c before clustering occurs. Here, we
represent the structure of the unlabeled data sets
as a Reordered Dissimilarity Image (RDI), where
pair wise dissimilarity information about a data set
including n objects is represented as nxn image.
RDI is generated using VAT (Visual Assessment of
Cluster tendency), RDI highlights potential clusters
as a set of dark blocks along the diagonal of the
image. So, number of clusters can be easily
estimated using the number of dark blocks across
the diagonal. We develop a new method called
Extended Dark Block Extraction (EDBE) for
counting the number of clusters formed along the
diagonal of the RDI. EDBE method combines
several image and signal processing techniques.
General Terms: Data Mining, Image Processing,
Artificial Intelligence.
KEYWORDS
Clustering, Cluster Tendency, Reordered
Dissimilarity Image, VAT, C-Means Clustering.
1. INTRODUCTION
The main Objective of our work Estimating the
number of clusters in unlabeled data sets is to
determine the number of clusters c prior to
clustering. Many clustering algorithms require
number of clusters c as an input parameter, so
the quality of clusters is largely dependent on the
estimation of the value c. Most methods are post
clustering measures of cluster validity i.e. they
attempt to choose the best partition from a set of
alternative partitions. In contrast, tendency
assessment attempts to estimate c before
clustering occurs. Our focus is on preclustering
tendency assessment.
The existing technique for preclustering
assessment of cluster tendency is Cluster Count
Extraction (CCE). The results obtained from this
are less accurate and less reliable. It does not
concentrate on the perplexing and overlap issues.
Its efficiency is also doubted. Hence we are
introducing a new technique in our work. Our
work mainly includes two algorithms, i.e. Visual
Assessment of Cluster Tendency (VAT) and
Extended Dark Block Extraction (EDBE). Here, we
initially concentrate on representation of
structure in unlabeled data in an image format.
Then for that image VAT algorithm is applied, and
then for the output of VAT, we apply EDBE
algorithm, there by generating the valid number
of peaks (i.e. number of clusters). Pair wise
dissimilarity information of a dataset including n
objects is depicted as an n*n image, where the
objects are potentially reordered so that the
resultant image is better able to highlight the
potential cluster structure of the data. The
intensity of each pixel in the RDI corresponds to
the dissimilarity between the pair of objects
addressed by the row and column of the pixel. A
useful RDI highlights potential clusters as a set
of dark blocks along the diagonal of the image,
corresponding to sets of objects with low
dissimilarity.
This dissimilarity matrix generated will be
provided as input to the VAT algorithm. RDI
NICE-2010
Acharya Institute of Technology, Bangalore-560090 50
(Reordered Dissimilarity Image) that portrays a
potential cluster structure from the pair wise
dissimilarity matrix of the data is created using
VAT. Then, sequential image processing
operations (region segmentation, directional
morphological filtering, and distance
transformation) are used to segment the regions
of interest in the RDI and to convert the filtered
image into a distance-transformed image. Finally,
we project the transformed image onto the
diagonal axis of the RDI, which yields a one-
dimensional signal, from which we can extract the
(potential) number of clusters in the data set
using sequential signal processing operations like
average smoothing and peak detection. The peaks
and valleys are found using peak detection
techniques from the projected signal. These peaks
and valleys are made to satisfy certain conditions.
Only the peaks which satisfy the given condition
will be considered as valid peaks. The number of
valid peaks provides the number of clusters that
can be formed from the unlabeled data sets. The
proposed method is easy to understand and
implement, and thereby encouraging results are
achieved.
2. RELATED WORK
Visual methods for cluster tendency assessment
for various data analysis problems have been
widely studied [10], [5], [9]. For data that can be
projected onto a 2D Euclidean space (which are
commonly depicted with a scatter plot), direct
observations can provide a good insight on the
value of c. Apparently, Ling [1] first automated the
creation of the RDI in
1973 with an algorithm called SHADE, which was
used after the application of the complete linkage
hierarchical clustering
scheme and served as an alternative to visual
displays of hierarchically nested clusters via the
standard dendrogram. Since then, there have been
many studies of the best method for reordering
and for the use of RDIs in clustering. Two general
approaches have emerged, depending on whether
the RDI is viewed before or after clustering. Most
RDIs built for viewing prior to clustering use
algorithms very similar in flavor to single-linkage
to reorder the input dissimilarities, and the RDI is
viewed as a visual aid to tendency assessment.
This is the problem addressed by our new DBE
algorithm, which uses the VAT algorithm of
Bezdek and Hathaway [2] to find RDIs. VAT is
related but not identical to single-linkage
clustering; see [11] for a detailed analysis of this
aspect of VAT. Several algorithms extend VAT for
related assessment problems. The bigVAT [3] and
sVAT [4] offered different ways to approximate
the VAT RDI for very large data sets. The coVAT
[6] extended the idea of RDIs to rectangular
dissimilarity data to enable tendency assessment
for each of the four co-clustering problems
associated with such data.
2.1. Review of VAT
The visual approach for assessing cluster
tendency introduced here can be used in all cases
involving numerical data. It is both convenient
and expected that new methods in clustering have
a catchy acronym. Consequently, we call this new
tool VAT (visual assessment of tendency). The VAT
approach presents pair wise dissimilarity
information about the set of objects O = {o1 on}
as a square digital image with n2 pixels, after the
objects are suitably reordered so that the image is
better able to highlight potential cluster structure.
To go further into the VAT approach requires
some additional background on the types of data
typically available to describe the set O = {o1
on}.
There are two common data representations of O
upon which clustering can be based. When each
object in O is represented by a (column) vector x
in s , the set X = {x1xn} s is called an object
data representation of O. The VAT tool is widely
applicable because it displays a reordered form of
dissimilarity data, which itself can always be
obtained from the original data for O. If the
original data consists of a matrix of pair wise
(symmetric) similarities S = [Sij], then
dissimilarities can be obtained through several
simple transformations.
For example, we can take Rij = Smax Sij, where
Smax denotes the largest similarity value. If the
original data set consists of object data X = {x1
xn} s, then Rij can be computed as Rij = xi x j,
using any convenient norm on s, the VAT
approach is applicable to virtually all numerical
data sets
NICE-2010
Acharya Institute of Technology, Bangalore-560090 51
Fig. 1a is a scatter plot of n 3,000 data points in
R2, These data points were converted to a 3,000 *
3,000 dissimilarity matrix D by computing the
Euclidean distance between each pair of points.
The five visually apparent clusters in Fig. 1a are
reflected by the five distinct dark blocks along the
main diagonal in Fig. 1c, which is the VAT image of
the data after reordering. Compared with Fig. 1b,
which is the image of dissimilarities D in original
input order, we can say that reordering is
necessary to reveal the underlying cluster
structure of the data. The reordering method of
VAT is summarized in Table 1
3. VAT ALGORITHM
Step 1) A dissimilarity matrix m of size n*n is
generated from the input dataset S, where n is
the size of S; //initialization
Step 2) set K{1,2,3.n},
IJ{},P[]{0,0,00};
Step 3) select ( i , j) argmax (mpq ) such that
(p,q) K and set P[1]i; I{i},
JK-{i};
Step 4) for r2,3n
Select(i, j) argmin (mpq) and set P[r] j, IIU{j},
JJ-{j}
Next r
Step 5) Obtain the ordered dissimilarity matrix R
using the ordering array P as
. Rij = mp(i)p(j) for 1<=i, j<=n.
Step 6) Display the Reordered Dissimilarity
Image.
4. EDBE ALGORITHM
The existing system for automatically determining
the number of clusters in unlabeled data sets is
cluster count extraction.
Because of its limitations like perplexing, and its
inability in histogram overlapping, we are moving
on to a new technique. The proposed system is
Extended Dark Block Extraction, which is nearly
a parameter free method developed to
automatically determine the number of clusters in
unlabeled datasets. In short, EDBE is an algorithm
that counts the dark blocks along the diagonal of a
RDI.
EDBE algorithm mainly includes four major steps:
Dissimilarity Transformation and Image
segmentation.
Directional Morphological filtering of binary
image.
Distance transform and diagonal projection of
filtered image.
Detection of major peaks and valleys in the
projected signal
4.1 Dissimilarity transformation and Image
Segmentation (Steps 1-3):
Because information about possible cluster
structure in the data is embodied in the dark
blocks in the RDI, an important preprocessing
step is image thresholding to extract the regions
of interest. Choosing a threshold around the
first mode is thus ideal for image segmentation.
Otsus algorithm [7], which maximizes the
between class variance, has been widely used in
image processing for automatically choosing a
global threshold.
f (t) = 1 exp( -t / )
EDBE ALGORITHM
Step 1) Find the threshold value from m using
otsus algorithm.
Step 2) Transform m in to new dissimilarity
matrix m1 with m1ij = 1- exp(-m/)
Step 3) Form an RDI image I
1
using the previous
module.
Step 4) Threshold I
1
to obtain a binary image I
2
using algorithm of otsu.
Step 5) Filter I
2
using morphological operations
to obtain a filtered binary image I
3
.
Step 6) Perform a distance transform on I
3
to
obtain a gray scale image I
4
and scale the
pixel values to [0, 1].
Step 7) Project the pixel values of the image on to
the main diagonal axis of I
4
to form a
projection signal H
1
Step 8) Smooth the signal H
1
to obtain the
filtered signal H
2
by an average filter.
NICE-2010
Acharya Institute of Technology, Bangalore-560090 52
Step 9) Compute the first order derivative of H
2
to obtain H
3
.
Step 10) Find peak position p
i
and valley
positions v
j
in H
3
.
Step 11) Select valid peaks by considering some
conditions. Number of valid peaks gives
number of clusters.
Step 12) Put the number of clusters into C-Means
Clustering Algorithm and gives very good
accuracy.
This does not affect the reordering by VAT but
changes the histogram of dissimilarities. From the
histogram of D0, we use Otsus algorithm again to
obtain a new threshold to convert the VAT image
shown in fig 2a into a binary image shown in
fig.2b by
Iij
2
= 1, if Iij
2
>
Iij
2 = 0,
otherwise.
It can be seen that the segmentation result after
transformation is far better than that before
transformation.
Directional morphological filtering of binary
image
(Step 4):
To make the segmented image clearer, especially
for the cases in which the degree of overlap
between clusters is large, we use morphological
operations [8] to perform binary image filtering.
Morphological filtering is one type of processing
in which the spatial form or structure of objects
within an image is modified. Dilation and erosion
are two fundamental morphological operations.
The former usually causes objects to grow in size,
while the latter causes objects to shrink. The
morphologically filtered image is as shown in the
fig.3a
Fig 3a: Morphologically filtered Image
4.2. Distance Transform and diagonal
projection of image (Steps 5-6):
In order to convert the morphologically filtered
image into an informative one that clearly shows
the dark block structure information; we need to
consider the values of pixels that are
along or off the main diagonal axis of the image.
First, we perform a DT of the binary image to
obtain a new gray-scale image as shown in the
fig.3b A Distance Transform is a form of
representation of a digital image, which converts
a binary image to a gray-scale image in which the
value of each pixel is the distance from the pixel to
the nearest nonzero pixel in the binary image
fig.3b.
There are several different DTs depending upon
which distance metric is being used to determine
the distance between pixels. We use the Euclidean
distance. After the DT,
NICE-2010
Acharya Institute of Technology, Bangalore-560090 53
we project all pixel values of the DT image onto
the main diagonal axis to obtain a projection
signal as shown in the fig. 3c.
Fig.3b Distance Transformed Image(I
4
)
Fig.3c Diagonal projection signal from (I
4
)
4.3. Detection of major peaks and valleys in the
projected signal (Steps 7-10):
The number of dark blocks in any RDI is
equivalent to the number of major peaks in the
projection signal H
1
. We perform the detection of
peaks and valleys to estimate the (cluster)
number c, based on the first-order derivative of
the projection signal. Although the projection
signal H1
seems to be very smooth, we require
further smoothing to reduce possible false
detections due to noise in the signal. Here, we use
a simple average filter h to filter the projection
signal,
i.e., H
(2)
= h * H
(1),
where * means linear
convolution (see Fig. 3c), and the average filter h
has length l2= 2**n.
Fig.3d First Order derivative signal
After that, the process of peak and valley
detection is performed in a from-rough-to-fine
manner. It is well known that the peaks and
valleys of a signal usually correspond to zero-
crossing points in its first-order derivative as
shown in the fig.3d. Accordingly, we can find the
initial sets of peaks pi and valleys vj by finding the
corresponding from positive-to-negative zero-
crossing points and from negative-to-positive
zero-crossing points. To further remove minor
false peaks, we use a size filter to remove
relatively small valleys by validating the width
between each two neighboring valleys.
That is, the peak pi within the two neighboring
valleys will be kept as a meaningful major peak
if V(k+1) - V(k) > l3
V(k) < P(i) < V(k+1) , where l3=2n
Finally, we determine the number of dark blocks
in the RDI (and, hopefully, the number of clusters
c in the unlabeled data) as the number of resulting
major peaks.
5. CONCLUSION
This paper investigates a nearly parameter-free
method for automatically estimating the number
of clusters in unlabeled data sets. The only user
defined parameter that must be chosen controls
the filter size. It is relatively easy to make a
realistic (and useful) choice for , since it
essentially specifies the smallest cardinality of a
cluster relative to the number of objects in the
data. Cluster number should be EDBE that will
NICE-2010
Acharya Institute of Technology, Bangalore-560090 54
probably reach its useful limit when the RDI
formed by any reordering of D is not from a well
structured dissimilarity matrix. In our
experiments, we used the simple Euclidean
distance to compute pair wise dissimilarities
when the input data are feature vectors. The
Euclidean distance may not be suitable for high
dimensional or complex data valleys (such as
wavelet-based multi-resolution analysis). EDBE
provides an initial estimation of the cluster
number, thus avoiding the requirement of
repeatedly running a clustering algorithm
multiple times over a wide range of c in an
attempt to find useful clusters. In this way, EDBE
compares favorably to post clustering validation
methods in computational efficiency. It is noted
that EDBE does not eliminate the need for cluster
validity, but it simply improves the probability of
success. A possible extension of this work
concerns the initialization of the fuzzy post
clustering algorithm for object data clustering. It
should not be too hard to find an approximate
center sample for each meaningful cluster from
any well structured RDI.
6. REFERENCES
[1] R.F. Ling, Comm. ACM, vol. 16, pp. 355-361,
1973, A Computer Generated Aid for Cluster
Analysis,
[2] J. Huband, J.C. Bezdek, and R. Hathaway,
Pattern Recognition, vol. 38, no. 11, pp. 1875-
1886, 2005, bigVAT: Visual Assessment of
Cluster Tendency for Large Data Sets.
[3] R. Hathaway, J.C. Bezdek, and J. Huband,
Pattern Recognition, vol. 39, pp. 1315-1324,
2006, Scalable Visual Assessment of Cluster
Tendency.
[4] W.S. Cleveland, Visualizing Data. Hobart Press,
1993. [6] J.C. Bezdek, R.J. Hathaway, and J.
Huband, IEEE Trans. Fuzzy Systems, vol. 15,
no. 5, pp. 890-903, 2007, Visual Assessment
of Clustering Tendency for Rectangular
Dissimilarity Matrices.
[5] R.C. Gonzalez and R.E. Woods, Prentice Hall,
2002, Digital Image Processing.
[6] I. Dhillon, D. Modha, and W. Spangler, Proc.
30th Symp. Interface: Computing Science and
Statistics, 1998, Visualizing Class Structure of
Multidimensional Data.
[7] T. Tran-Luu, PhD dissertation, Univ. of
Maryland, College Park, 1996, Mathematical
Concepts and Novel Heuristic Methods for
Data Clustering and Visualization.
[8] J.C. Bezdek and R. Hathaway, Proc. Intl Joint
Conf. Neural Networks (IJCNN 02), pp. 2225-
2230, 2002, VAT: A Tool for Visual
Assessment of (Cluster) Tendency.
[9] Liang Wang, Christopher Leckie, Kotagiri
Ramamohanarao, and James Bezdek, Fellow,
IEEE-MARCH 2009, Automatically
Determining the Number of Clusters in
Unlabeled Data Sets.
NICE-2010
Acharya Institute of Technology, Bangalore-560090 55
PREDICTION BASED LOSSLESS COMPRESSION SCHEME FOR BAYER COLOR
FILTER ARRAY IMAGE
Khajavali Shaik
1
, Mr.M.N.A.Siddiqui
2
, Mrs. C H. Hima Bindu.
3
P.G student in ECE Department
1
, P.G student in ECE Department
2
, Sr.Associate Prof in Dept of ECE
3
QIS College of Engg & Technology, Ongole,Prakasam(Dt),A.P.
Khajavali.shaik440@gmail.com
1
,
Siddiqui.nadeem07@gmail.com
2
, hb.muvvala@gmail.com
3
_____________________________________________________________________________________________________________________________
ABSTRACT
In most digital cameras Bayer color filter
array images captured and demosaicing is
generally carried out before compression. Recently
it was compression first scheme o u t perform the
conven-tional demosaicing first schemes in terms
of output image quality. An ef f i ci ent prediction
based lossless compression scheme for Bayer filter
color images proposed
INTRODUCTION
BAYER COLOR FILTER ARRAY
A Bayer Filter color array usually coated over the
sensors in these cameras to record only one of
the three colors components at each pixel
location. The resultant image is referred to as a
CFA image.
Fig: Bayer Patter has Red sample in center
Fig shows the Bayer Patter has Red sample in
center, compressed for storage. Then it was
inefficient in a way the demosaicing process
always introduce some redundancy which
should eventually be removed in the following
compression step. We do the compression before
demosaicing digital cameras can have a simpler
design and low power consumption as
computationally heavy process like demosaicing
can be carried in an offline powerful personal
computer. This motivates the demand of CFA
image compression schemes.
Fig: single sensor camera imaging chain (a)
demosaicing and (b) Compression
PRESENT SCHEMES USED
There are different schemes present in the
market such as
Lossy compression schemes JPEG2000
So now we have to look the drawbacks of present
methods.
Lossy schemes compress a CFA Image by
discarding its visually redundant information.
This scheme visually yields a higher
compression ratio as compared with the
lossless schemes.
JPEG-2000 is used to encode a CFA image but
only a fair performance can be attained.
JPEG-2000 is very expensive method to
compress the images.
PROPOSED SCHEME
A Prediction based lossless CFA compression
scheme is proposed. It divides a CFA images into
two sub-images:
(a) A green sub-image which contains all green
samples of the CFA image
(b) Non-green sub image which contains the red
and blue samples in the CFA image.
This system is mainly consists of two parts
(a) Encoder
(b) Decoder
ENCODER:
Fig: Structure of proposed scheme
Green Sub image is coded first and the Non
green Sub image follows based on green sub
image as reference and To reduce the spectral
NICE-2010
Acharya Institute of Technology, Bangalore-560090 56
redundancy, the non green sub image is
processed in the color difference domain
whereas the green sub image is processed in the
intensity domain as a reference for the color
difference content of the non green sub image.
Both sub images are processed in raster scan
sequence with context matching based
prediction technique to remove the spatial
dependency. The prediction residue planes of
the two sub images are then entropy encoded
sequentially with our proposed realization
scheme of adaptive Rice code.
WORKING OF THE SCHEME
This proposed scheme is mainly working on
Prediction on the green plane and Prediction on
the Non-green plane.
PREDICTION ON THE GREEN PLANE:
As the green plane is raster scanned during the
prediction and all prediction errors are
recorded. Now processing a particular green
plane the four nearest processed neighboring
samples of g (i,j) form a candidate set.
We can find the directions associated with the
green pixels it need some process.
Fig: Four possible directions associated with
green pixel
Let g(mk,nk) g(i,j) for k=1,2,3,4 be the four
ranked candidates of sample g(i,j) (Sg(i,j),
Sg(mu,nu)) <= D(Sg(i,j), Sg(mv,nv) ) for
1<=u<=v<=4
If the directions of g(i,j) is identical to the
directions of all green samples in Sg(i,j), pixel (i,j)
will be considered in a homogenous region and
prediction of g(i,j) is
i.e. {w1,w2,w3,w4}={1,0,0,0}
Else the g(i,j) is in heterogenous region and
predicted value of g(i,j) is
i.e. {w1,w2,w3,w4}={5/8,2/8,1/8,0}
FLOW CHART FOR PREDICTION ON THE
GREEN PLANE
ADAPTIVE COLOR DIFFERENCE
ESTIMATION FOR NON GREEN PLANE
When compressing the non green color plane,
color difference information is exploited to
remove the color spectral dependency.
Let c(m,n) be the intensity value at a non green
sampling position(m,n). Green-Red(Green-Blue)
color difference of pixel (m,n) is
d(m,n)=g(m,n)-c(m,n)
g(m,n) estimated green component intensity
value
g(m,n)=round(( H*Gv+V*GH)/(H+ V))
GH=(g(m,n-1)+g(m,n+1))/2 and
Gv=(g(m+1,n)+g(m+1,n))/2
NICE-2010
Acharya Institute of Technology, Bangalore-560090 57
PREDICTION ON THE NON GREEN PLANE
Color difference prediction of a non green sample
c(i,j) with color difference value d(i,j) is
Where k is predictor coefficient
d(mk,nk) is kth ranked candidate in c(i,j)
COMPRESSION SCHEME
The prediction Error of pixel (i, j) in the CFA
image, say e(i, j) is given by
where g(i, j), d(i, j) are respectively, the real green
sample value and the color difference value of
pixel (i, j)
The error residue e(i, j) is then mapped to a
nonnegative integer as follows to reshape its
value distribution to an exponential one from a
Laplacian one.
The E(i, j) s from the green sub-image are raster
scanned and coded with Rice code first. Rice code
is employed to code E(i, j) because of its
simplicity and high efficiency in handling
exponentially distributed sources When
Rice code is used, each mapped Residue E(i, j) is
split into a Quotient Q
Quotient and Remainder are then saved for
storage and transmission.
The Length of code word used for representing
E(i, j) is k dependent and is given by
Parameter k is critical to the compression
performance as it determines the code length of
E(i, j)
Optimal parameter K is given by
For a geometric source with distribution
parameter color spaces I As long as is known,
parameter , and, hence, the optimal coding
parameter k for the whole source can be
determined easily.
is estimated adaptively in course of Encoding
DECODING PROCESS:
NICE-2010
Acharya Institute of Technology, Bangalore-560090 58
Decoding Process is just reverse process of
Encoding. Green Sub-image is decoded first and
then the non-green sub-image is decoded with
the decoded green sub-image as a reference.
Original CFA Image is then reconstructed by
combining the two sub-
EXPERIMENT RESULTS
From the above fig, it shows that = 1 can
provide a good compression performance. We
assume the prediction residue is a local
variable and estimate the mean of its value
distribution adaptively. The divisor used to
generate the Rice code is then adjusted
accordingly so as to improve the efficiency of Rice
code.
COMPRESSION PERFORMANCE
Simulations were carried out to evaluate the
performance of proposed compression
scheme. 24-bit color images of size 512*768
were sub- sampled according to the Bayer
pattern to form 8 bit testing CFA images. These
Images are directly coded by the proposed
compression scheme for evaluation. Some
representative Lossless compression
schemes such as JPEG-LS, JPEG 2000(lossless
mode) and LCMI were used for comparison of
Results.
ADVANTAGES OF PROPOSED METHOD
We can reduce the spectral redundancy mean
time and also can get high quality image.
Reducing the sensors in digital cameras from
3 to 1. Low complexity to design. Compare with
JPEG2000 it gives better performance.
SIMULATION RESULTS
EXPERIMENTAL RESULTS
CONCLUSION
CFA image encodes the sub-image separately
with predictive coding Lossless prediction is
carried out in the intensity domain for the
green. While it is carried out in the color
difference domain for the non green.
NICE-2010
Acharya Institute of Technology, Bangalore-560090 59
REFERENCES
[1] S. Banks, Signal Processing, Image Processing
and Pattern Recognition.
Englewood Cliffs, NJ: Prentice-Hall, 1990.
[2] S. P. Lloyd, Least squares quantization in
PCM, IEEE Trans. Inf.
Theory, vol. IT-28, no. 2, pp. 129136, Mar. 1982.
[3] P. Berkhin, Survey of clustering data mining
techniques, Accrue
Software, San Jose, CA, 2002.
[4] J. Besag, On the statistical analysis of dirty
pictures, J. Roy. Statist.
Soc. B, vol. 48, pp. 259302, 1986.
[5] D. Comaniciu and P. Meer, Mean shift: A
robust approach toward
feature space analysis, IEEE Trans. Pattern Anal.
Mach. Intell., vol.
24, no. 5, pp. 603619, May 2002.
[6] J. Shi and J. Malik, Normalized cuts and image
segmentation, IEEE
Trans. Pattern Anal. Mach. Intell., vol. 22, no. 8, pp.
888905, Aug.
2000.
[7] P. Felzenszwalb and D. Huttenlocher, Efficient
graph-based image
segmentation, Int. J. Comput. Vis., vol. 59, pp. 167
181, 2004.
[8] S. Zhu and A. Yuille, Region competition:
Unifying snakes, region
growing, and Bayes/MDL for multiband image
segmentation, IEEE
Trans. Pattern Anal. Mach. Intell., vol. 18, no. 9, pp.
884900, Sep.
1996.
[9] M. Mignotte, C. Collet, P. Prez, and P.
Bouthemy, Sonar image
segmentation using a hierarchical MRF model,
IEEE Trans. Image
Process., vol. 9, no. 7, pp. 12161231, Jul. 2000.
[10] M. Mignotte, C. Collet, P. Prez, and P.
Bouthemy, Three-class Markovian
segmentation of high resolution sonar images,
Comput. Vis.
Image Understand., vol. 76, no. 3, pp. 191204,
1999.
[11] F. Destrempes, J.-F. Angers, and M. Mignotte,
Fusion of hidden
Markov random field models and its Bayesian
estimation, IEEE
Trans. Image Process., vol. 15, no. 10, pp. 2920
2935, Oct. 2006.
[12] Z. Kato, T. C. Pong, and G. Q. Song,
Unsupervised segmentation of
color textured images using a multi-layer MRF
model, in Proc. Int.
Conf. Image Processing, Barcelona, Spain, Sep.
2003, pp. 961964.
[13] P. Prez, C. Hue, J. Vermaak, and M. Gangnet,
Color-based probabilistic
tracking, in Proc. Eur. Conf. Computer Vision,
Copenhagen,
Denmark, Jun. 2002, pp. 661675.
[14] J. B. Martinkauppi, M. N. Soriano, and M. H.
Laaksonen, Behavior
of skin color under varying illumination seen by
different cameras at
different color spaces, in Proc. SPIE, Machine
Vision Applications in
Industrial Inspection IX, San Jose, CA, Jan. 2001, pp.
102113.
[15] J.-P. Braquelaire and L. Brun, Comparison
and optimization of
methods of color image quantization, IEEE Trans.
Image Process.,
vol. 6, no. 7, pp. 10481952, Jul. 1997.
[16] H. Stokman and T. Gevers, Selection and
fusion of color models for
NICE-2010
Acharya Institute of Technology, Bangalore-560090 60
VIDEO COMPRESSION USING ACC-JPEG
Revathi Amisetty, M.Doss
Asst. professor, JNTU, Anantapur.
_____________________________________________________________________________________________________________________________
ABSTRACT
The high correlation between successive
frames has not been exploited by the present
compression techniques. In this paper, we present a
new video compression approach which tends to
hard exploit the temporal redundancy in the video
frames to improve compression efficiency with
minimum processing complexity. The accordion
representation of video consists on 3D to 2D
transformation of the video frames that allows
exploring the temporal redundancy of the video
using 2D transforms and avoiding the
computationally demanding motion compensation
step. This transformation turns the spatial
temporal correlation of the video into high spatial
correlation. Indeed, this technique transforms each
group of pictures to one picture eventually with
high spatial correlation which is compressed as a
still image by JPEG. Thus, the de-correlation of the
resulting pictures by the DCT makes efficient
energy compaction, and therefore produces a high
video compression ratio. This method proved to be
more efficient especially in high bit rate and with
slow motion video.
INTRODUCTION
The objective of video coding in most video
applications is to reduce the amount of video data
for storing or transmission purposes without
affecting the visual quality. The desired video
performances depend on applications
requirements, in terms of quality, disks capacity
and bandwidth. For portable digital video
applications, highly-integrated real-time video
compression and decompression solutions are
more and more required. Actually, motion
estimation based encoders are the most widely
used in video compression. Such encoders
exploits inter frame correlation to provide more
efficient compression. However, Motion
estimation process is computationally intensive;
its real time implementation is difficult and costly.
This is why motion-based video coding standard
MPEG was primarily developed for stored video
applications, where the encoding process is
typically carried out off-line on powerful
computers. So it is less appropriate to be
implemented as a real-time compression process
for a portable recording or communication device
(video surveillance camera and fully digital video
cameras). In these applications, efficient low
cost/complexity implementation is the most
critical issue. Thus, researches turned towards the
design of new coders more adapted to new video
applications requirements. This led some
researchers to look for the exploitation of 3D
transforms in order to exploit temporal
redundancy. Coder based on 3D transform
produces video compression ratio which is close
to the motion estimation based coding one with
less complex processing. The 3d transform based
video compression methods treat the
redundancies in the 3D video signal in the same
way, which can reduce the efficiency of these
methods as pixel's values variation in spatial or
temporal dimensions is not uniform and so,
redundancy has not the same pertinence. Often
the temporal redundancies are more relevant
than spatial one. It is possible to achieve more
efficient compression by exploiting more and
more the redundancies in the temporal domain;
this is the basic purpose of the proposed method.
The proposed method consists on projecting
temporal redundancy of each group of pictures
into spatial domain to be combined with spatial
redundancy in one representation with high
spatial correlation. The obtained representation
will be compressed as still image with JPEG coder.
PROPOSED APPROACH
The basic idea is to represent video data with high
correlated form. Thus, we have to exploit both
temporal and spatial redundancies in video signal.
The input of our encoder is so called video cube,
which is made up of a number of frames. This
cube will be decomposed into temporal frames
which will be gathered into one frame (2
dimensions). The final step consists of coding the
obtained frame. In the following, we detail the
method design steps.
A. Hypothesis
Many experiences had proved that the variation of
the 3D video signal is much less in the temporal
dimension than the spatial one. Thus, pixels, in 3D
video signal, are more correlated in temporal
domain than in spatial one [3]; this could be
traduced by the following expression: for one
reference pixel I(x;y;t) where:
1. I: pixel intensity value
2. x; y: space coordinate of the pixel
NICE-2010
Acharya Institute of Technology, Bangalore-560090 61
3. t: time (video instance)
We could have generally:
I(x; y;t)-I(x; y; t+1) < I(x;y;t) I(x+1; y;t) (1)
This assumption will be the basis of the proposed
method where we will try to put pixels - which
have a very high temporal correlation - in spatial
adjacency.
Accordion based representation
To exploit this succeeding assumption, we start by
carrying out a temporal decomposition of the 3D
video signal, the shows temporal and spatial
decomposition of one 8X8X8video cube:
Frames obtained following the temporal
decomposition will be called temporal frames.
These latter are formed by gathering the video
cube pixels which have the same column rank.
According to the mentioned assumption, these
frames have a stronger correlation compared to
spatial frames. To increase correlation in
Accordion Representation we reverse the
direction of event frames. Figure 4 illustrates the
principle of this representation.
Thus, the Accordion representation is obtained
as following: first, we start by carrying out a
temporal decomposition of the video 3D. Then,
the event temporal frames will be turned over
horizontally (Mirror effect). The last step consists
of frames successive projecting on a 2D plan
further called "IACC" frame. The Accordion
representation tends to put in spatial adjacency
the pixels having the same coordinates in the
different frames of the video cube. This
representation transform temporal correlation in
the 3D original video source into a high spatial
correlation in the 2D representation ("IACC"). The
goal of turning over horizontally the event
temporal frames is to more exploit the spatial
correlation of the video cube frames extremities.
In this way, Accordion representation also
minimizes the distances between the pixels
correlated in the source. That's could be clearer in
figure 3:
Figure 4 shows the strong correlation obtained in
the Accordion representation made of 4 frames
which show motion of Rhinosaur.
Accordion analytic representation
The Accordion representation is obtained
following a process having as input the GOP
frames(I 1..N) and has as output the resulting
frame IACC. The inverse process has as input the
IACC frame and as output the coded frames (I
1..N). The analysis of these two processes leads to
the following algorithms:
The algorithm 1 describes how to make Accordion
representation (labeled ACC), The algorithm 2
represents the process inverse (labeled ACC
-1
).
Let us note that:
1) L and H are respectively the length and the
height of the video source frames.
2) NR is the number of frames of a GOP.
Algorithm 1: Algorithm of ACC:
1: for x from 0 to (L * N)- 1 do
2: for y from 0 to H - 1 do
3: if ((x div N) mod 2) != 0 then
4: n=(N-1) - (x mod N)
5: else
6: n=x mod N
7: end if
8: IACC(x; y)= In (x div N;y)
9: end for
10: end for
Algorithm 2 : Algorithm of ACC
-1
:
1: for n from 0 to N - 1 do
2: for x from 0 to L - 1 do
3: for y from 0 to H - 1 do
4: if (x mod 2) != 0 then
NICE-2010
Acharya Institute of Technology, Bangalore-560090 62
5: XACC= (N -1) - n (x*N)
6: else
7: XACC= n(x * N)
8: end if
9: In(x;y)=IACC(XACC; y)
10: end for
11: end for
12: end for
3) IACC(x; y) is the intensity of the pixel which is
situated in "IACC" frame with the co-ordinates
(x,y) according to Accordion representation
repair.
4) In(x; y) is the intensity of pixel situated in the
Nth frame in original video source. We can also
present the Accordion Representation With the
following formulas:
ACC formulas:
IACC=In(xdivN;y) (2)
with n=((x div N) mod2)(N-1) + 1-2((x div N) mod
2)(x mod N)
ACC inverse formulas:
In(x;y)=IACC(XACC;y) (3)
with XACC = ((x div N) mod 2)(N-1) + n(1-2('x div
N) mod 2))+x
In the following, we will present the diagram of
coding based on the .Accordion representation.
Diagram of coding ACC -JPEG
ACC-JPEG Coding is proceded as follows:
1) Decomposition of the video in groups of frames
2) Accordion Representation of the GOP.
3)Decomposition of the resulting "IACC" frame
into 8x8 blocks.
4) For each 8x8 block:
Discrete cosine Transformation (DCT).
Quantification of the obtained coefficients.
Course in Zigzag of the quantized coefficients.
Entropic Coding of the coefficients (RLE,
Huffman)
ACC - JPEG FEATURE ANALYSIS
The proposed method presents several
advantages:-
Random Access: 3D transform and motion
estimation based video compression methods
require all the frames of the GOP to allow the
random access to different frames. However, the
proposed method allows the random frame
access. The ACC formula makes it possible to code
and/or decode a well defined zone of the GOP
(Partial coding). As conclusion, we can state that
the ACC - JPEG is very efficient for scenes with a
translator character [9], or with slow motion,
especially without change of video plan.
Furthermore, ACC - JPEG seems to be well
adapted to embedded or portable video devices
such as the IP cameras thanks to its flexibility and
its operating simplicity.
symmetry: On the contrary of coding schemes
based on motion estimation and compensation
whose coding is more complex than decoding, the
proposed encoder and decoder are symmetric
with almost identical structure and complexity,
which facilitates their joint implementation.
Simplicity: The proposed method transform the
3D features to 2D ones, which enormously reduce
the processing complexity. Moreover, the
complexity is independent of the compression
ratio and motions.
Objectivity: Unlike 3D methods that treat
temporal and spatial redundancies in the same
way, the proposed method is rather selective., it
exploits the temporal redundancies more than the
space redundancies; what is more objective and
more efficient.
Flexibility: The parameters of the ACC--JPEG
offer a flexibility that makes it possible to be
adapted to different requirements of video
applications: The latency time, the compression
ratio and the size of required memory depend on
the value of the NR parameter. Indeed, by
increasing the value of NR, the compression ratio,
the latency time and the reserved memory
increase. This parameter allows to optimize the
Compression/ Quality compromise while taking
in consideration memory and latency constraints.
CONCLUSION
In this research, the proposed method could open
new horizons in video compression domain; it
strongly exploits temporal redundancy with the
minimum of processing complexity which
facilitates its implementation in video embedded
systems. This new video compression method
which exploits objectively the temporal
redundancy. It presents some useful functions and
features which can be exploited in some domains
as video surveillance. In high bit rate, it gives the
best compromise between quality and complexity.
It provides better performance than MJPEG and
MJPEG2000 almost in different bit rate values.
Over 2000kb/s bit rate values our compression
method performance becomes comparable to the
NICE-2010
Acharya Institute of Technology, Bangalore-560090 63
MPEG4 especially for low motion sequences.
There are various directions for future
investigations. First of all, we would like to
explore others possibilities of video
representation. Another direction could be to
combine Accordion representation with other
transformations such as wavelet transformation.
The latter allows a global processing on the whole
of the Accordion representation on the contrary of
the DCT which generally acts on blocks.
REFERENCES
[1] R. A.Burg, .A 3d-dct real-time video
compression system for low complexity
singlechip vlsi implementation,. in the Mobile
Multimedia Conf. (MoMuC), 2000.
[2] T. Fryza, Compression of Video Signals by 3D-
DCT Transform. Diploma thesis, Institute of Radio
Electronics, FEKT Brno University of Technology,
Czech Republic, 2002.
[3] E. Q. L. X. Zhou and Y. Chen, .Implementation of
h.264 decoder on general purpose processors
with media instructions, in SPIE Conf. on Image
and Video Communications and Processing, (Santa
Clara, CA), pp. 224.235, Jan 2003.
[4] M. B. T. Q. N. A. Molino, F. Vacca, Low
complexity video codec for mobile video
conferencing,. in Eur. Signal Processing Conf.
(EUSIPCO), (Vienna, Austria), pp. 665.668, Sept
2004.
[5] S. B. Gokturk and A. M. Aaron, .Applying 3d
methods to video for compression,. in Digital
Video Processing (EE392J) Projects Winter Quarter,
2002.
[6] T. Fryza, Compression of Video Signals by 3D-
DCT Transform. Diploma thesis, Institute of Radio
Electronics, FEKT Brno University of Technology,
Czech Republic, 2002.
[7] G. M.P. Servais, .Video compression using the
three dimensional discrete cosine transform,. in
Proc.COMSIG, pp. 27.32, 1997.
[6] R. A.Burg, .A 3d-dct real-time video
compression system for low complexity
singlechip vlsi implementation,. in the Mobile
Multimedia Conf. (MoMuC), 2000.
[7] A. N. N. T. R. K.R., .Discrete cosine transforms,.
in IEEE transactions on computing, pp. 90.93,
1974.
[8] T.Fryza and S.Hanus, .Video signals
transparency in consequence of 3d-dct transform,.
in Radioelektronika 2003 Conference Proceedings,
(Brno,Czech Republic), pp. 127.130, 2003.
[9] N. Boinovi and J. Konrad, .Motion analysis in
3d dct domain and its application to video coding,.
vol. 20, pp. 510.528, 2005.
[10] E. Y. Lam and J. W. Goodman, .A mathematical
analysis of the dct coef_cient distributions for
images,. vol. 9, pp. 1661.1666, 2000.
NICE-2010
Acharya Institute of Technology, Bangalore-560090 64
AN EFFICIENT IMAGE RETRIEVAL SYSTEM
Mrs. Seema Patil
1
, Mrs. Ashwini N
2
Lecturer, Sr Lecturer CSE Dept,TOCE
9900379841 theseema@gmail.com
1
9900525858 ashwinilaxman@gmail.com
2
_____________________________________________________________________________________________________________________________
ABSTRACT
An image retrieval system is a computer
system for browsing, searching and retrieving
images from a large database of digital images.
Most traditional and common methods of image
retrieval utilize some method of adding metadata
such as captioning, keywords, or descriptions to the
images so that retrieval can be performed over the
annotation words. Manual image annotation is
time-consuming, laborious and expensive .Another
method of image retrieval is content based image
retrieval which aims at avoiding the use of textual
descriptions and instead retrieves images based on
their visual similarity to a user-supplied query
image or user-specified image features. Content-
based image retrieval is the application of
computer vision to the image retrieval problem.
Content-based means that the search will analyze
the actual contents of the image. The term 'content'
in this context might refer to colors, textures, or
any other information that can be derived from the
image itself.
Keywords
Feature extraction, Feature matching, Histogram,
Texture
I. INTRODUCTION
Image retrieval which is based on
automatically extracted primitive feature such as
color. The need for image retrieval is to retrieve
images that are more appropriate, along with
multiple features for better retrieval accuracy.
Usually in search process using any search engine,
which is through text retrieval, which wont be so
accurate. So, we go for image retrieval system
using color feature. Image Retrieval system is also
known as content based image retrieval but we
consider all features of an image for example
Color, Texture and shape . CBIR is considered as
the process of retrieving desired images from
huge databases based on extracted features from
the image themselves without resorting to a
keyword [1]. Features are derived directly from
the images and they are extracted and analyzed
automatically by ,means of computer processing
[2]. Content based image retrieval is also know as
query by image content (QBIC) [3] and content-
based visual information retrieval (CBVIR).
Content-based means that the search makes use
of the contents of image themselves, rather than
relying on human-inputted metadata such as
captions or keywords. The similarity
measurements and the representation of the
visual features are two important issues in Image
Retrieval System.
Given a query image, with single /
multiple object present in it; mission of this work
is to retrieve similar kind of images from the
database based on the features extracted from the
query image. So a content- based search, for high
accuracy multiple features like color is
incorporated. Color feature extraction is done
through Global Color Histogram and Local Color
Histogram and Texture through Co- Occurrence &
Edge Frequency.
The visual features are classified into low
and high level features according to their
complexity and the use of semantics [l]. The use of
simple features like color or shape is not efficient
[4]. When retrieving images using combinations
of these- features there is a need for testing the
accuracy of these combinations and comparing
them with the single features based retrieval in
order to find the combinations that give the best
matches that enhance the performance of image
retrieval systems.[5]
In fact of some image retrieval systems
give good results for special cases of database
images. So one of the most important challenges
facing the evaluation of CBIR systems
performance is creating a common image
collection and obtaining relevance judgments [6].
On the other hand the use of similarity
measurements is very effective in image retrieval
systems. After extracting the required features
from images the retrieval process becomes the
measurement of similarity between the feature
vectors one for the query image and the other for
each database image.
Section II will present some of the related
work on image retrieval system. Section III
presents the proposed solution for image retrieval
NICE-2010
Acharya Institute of Technology, Bangalore-560090 65
system. Section IV shows the features
representations of an image . Section V shows the
feature extraction of an color and texture, .Section
VI presents the results for an image retrieval
system and finally with concluding remark
II.RELATED WORK
Mohamed A. Tahoun [7] suggest image
retrieval system as robust based content-based
image Retrieval System Using Multiple Features
Representations. The similarity measurements
and the representation of the visual features are
two important issues in Content-Based Image
Retrieval (CBIR). In this paper, he has compared
between the combination of wavelet-based
representations of the texture feature and the
color feature with and without using the color
layout feature. To represent the color information,
he has used Global Color Histogram - (GCH)
beside the color layout feature and with respect to
the texture information, we used Haar and
Daubechies wavelets. Based on some commonly
used Euclidean and Non-Euclidean similarity
measures, we tested different categories of
images and measured the retrieval accuracy when
combining such techniques..
Mari Partio [8] suggest most of the
existing mage retrieval systems perform
reasonably when using color features. However,
retrieval accuracy using shape or texture features
does not produce as good results. Therefore, this
thesis investigates different methods of
representing shape and texture in content-based
image retrieval. Later, when appropriate
segmentation algorithms are available some of
these methods could also be applied to video
object retrieval. The paper presents two
contributions: one is shape-based and the second
is texture-based retrieval method.
III. PROPOSED WORK FOR IMAGE
RETRIEVAL SYSTEM
A Definations
Image databases and collections can be
enormous in size, containing hundreds, thousands
or even millions of images. The conventional
method of image retrieval is searching for a
keyword that would match the descriptive
keyword assigned to the image by a human
categorizer. Currently under development, even
though several systems exist, is the retrieval of
images based on their color feature content called
Image Retrieval system using Color Feature.
While computationally expensive, the results are
far more accurate than conventional image
indexing. Hence, there exists a tradeoff between
accuracy and computational cost. This tradeoff
decreases as more efficient algorithms are utilized
and increased computational power becomes
inexpensive.
B Problem Specification
The problem involves entering an image
as a query into a software application that is
designed to employ image retrieval techniques in
extracting visual properties, and matching them.
This is done to retrieve images in the database
that are visually similar to the query image.
C Proposed Solution
The solution initially proposed was to
extract the primitive features of a query image
and compare them to those of database images.
The image features under consideration were
color. Thus, using matching and comparison
algorithms, the color features of one image are
compared and matched to the corresponding
features of another image. This comparison is
performed using color, distance metrics. In the
end, these metrics are performed one after
another, so as to retrieve database images that are
similar to the query. The similarity between
features was to be calculated using algorithms
used by well known CBIR systems such as QBIC.
For each specific feature there was a specific
algorithm for extraction and another for
matching.
D Block Diagram
Different types of user queries is inputted to
image retrieval system . With query by example
,the user searches with a query image and the
software finds images similar to it.[9]
In the image retrieval software each image is
stored in the database has its features extracted
and compared to the features of the query image .
It involves two steps.
NICE-2010
Acharya Institute of Technology, Bangalore-560090 66
Feature Extraction: The first step in image
retrieval is to extract color feature of the query
image.This extraction can be Average RGB or
color Histogram or Texture
Feature Matching
Once the features vectors are created, the
matching process performs the measuring of a
metric distance between the features vectors. And
results into best image retrieval.
Block diagram : When building an image database
,feature vectors from images are to be extracted
and then store the vectors in the database. Given
the query image its feature vectors are computed.
If the distance between feature vectors of the
query image and image in the database is small
enough the corresponding image in the database
is to considered as a match to the query.
Fig: Block Diagram of image retrieval systems
IV VISUAL FEATURES REPRESENTATIONS
One of the most important challenges
when building image based retrieval systems is
the choice and the representation of the visual
features [10]. Color is the most intuitive and
straight forward for the user while shape and
texture are also important visual amibutes but
there is no standard way to use them compared to
color for efficient image retrieval. Many content-
based image retrieval systems use color and
texture features [11].
In order to extract the selected features and index
the database images based on them, we used
Global Color Histogram (GCH) to extract the color
feature. Also in this section, the color layout
feature is extracted and the database images are
indexed based on the color layout feature.
A Color
Global Color Histogram (GCH) is the most
traditional way of describing the color attribute of
an image. It is constructed by computing the
normalized percentage of the color pixels in an
arrange corresponding to each color element. An
example of a true colored (RGB) image and the
corresponding histograms of each component are
displayed in Fig. 1.
To construct the color feature vector (its length is
256x3) for both the query image and all images in
the database, we identified the three-color
components (R, G, and B) and compute the
corresponding histograms of these components.
Fig: A colored image at the top, the three
components Red, Green and Blue in the middle,
and finally from left to right: the corresponding
histograms for Red, Green and Blue components
User
Interface
Query
Subsystem
Feature
Extractio
Feature Matching
Subsystem
Multi-
Modal
Feature Extraction
Clustering /
Query by
Example
Query by
Database Generation
Query and Retrieval
Multi-
Modal
Feature
Imported
NICE-2010
Acharya Institute of Technology, Bangalore-560090 67
The main method of representing color
information of images in CBIR systems is through
color histograms. A color histogram is a type of
bar graph, where each bar represents a particular
color of the color space being used.The color
space can RGB or HSV. The bars in a color
histogram are referred to as bins and they
represent the x-axis. The number of bins depends
on the number of colors there are in an image. The
y-axis denotes the number of pixels there are in
each bin. In other words how many pixels in an
image are of a particular color. Quantization in
terms of color histograms refers to the process of
reducing the number of bins by taking colors that
are very similar to each other and putting them in
the same bin.
There are two types of color histograms,
Global color histograms (GCHs) and Local color
histograms (LCHs). A GCH represents one whole
image with a single color histogram. An LCH
divides an image into fixed blocks and takes the
color histogram of each of those blocks. LCHs
contain more information about an image but are
computationally expensive when comparing
images.
The GCH is the traditional method for
color based image retrieval. However, it does not
include information concerning the color
distribution of the regions of an image. Thus when
comparing GCHs one might not always get a
proper result in terms of similarity of images.
B Global Color Histogram
The color histogram depicts color
distribution using a set of bins. Using the Global
Color Histogram(GCH), an image will be encoded
with its color histogram, and the distance
between two images will be determined by the
distance between their color histograms.
C Local Color Histogram
This approach (referred to as LCH)
includes information concerning the color
distribution of regions. The first step is to
segment the image into blocks and then to obtain
a color histogram for each block. An image will
then be represented by these histograms. When
comparing two images, we calculate the distance,
using their histograms, between a region in one
image and a region in same location in the other
image. The distance between the two images will
be determined by the sum of all these distances.
D Texture
Texture[14] measures look for visual patterns in
images and how they are spatially defined.
Textures are represented by texels (where A texel,
or texture element (also texture pixel) is the
fundamental unit of texture space used in
computer graphics) which are then placed into a
number of sets, depending on how many textures
are detected in the image. These sets not only
define the texture, but also where in the image the
texture is located. Texture is a difficult concept to
represent. The identification of specific textures in
an image is achieved primarily by modeling
texture as a two-dimensional gray level variation.
The relative brightness of pairs of pixels is
computed such that degree of contrast, regularity,
coarseness and directionality may be estimated.
Textures are represented by arrays of texels, just
as pictures are represented by arrays of pixels.
Texture through Co- Occurrence .Co-occurrence
matrix is a statistical method using second order
statistics to model the relationships between
pixels within the region by constructing Spatial
Gray Level Dependency (SGLD) matrices.
V. COLOR FEATURE EXTRACTION
A. Average RGB
Average RGB is to compute the average values in
R, G and B channel of each pixel in an image, and
use this as a descriptor of an image for
comparison purpose
Notation
I - an image
w - width of image I
h - height of image I
I(x,y) - the pixel of image I at row y,
column x
R(p), G(p), B(p) - the red, green and blue
color component of pixel p
ra, ga, ba - the average red, green and blue
component of image Ia
NICE-2010
Acharya Institute of Technology, Bangalore-560090 68
d(Ia,Ib) - the distance measure between
image Ia and Ib
Description Features
The following are 3 equations for
computing the average R, G, B component
of an image I
Here is the equation for distance measure
of image Ia and Ib, we use the weighted
Euclidean distance
The distance between two exact images
will be 0 and the distance between to
most dissimilar images (blank and white)
will be 1 or 255 depending on the range of
RGB is from 0-1 or 0-255.
B. Color Histogram
Color histograms are frequently used to
compare images. We discretize the colorspace of
the image such that there are n distinct
(discretized) colors. A color histogram H is a
vector (h1, h2, ..., hn), in which each bucket hj
contains the number of pixels of color j in the
image. Typically images are represented in the
RGB colorspace, and a few of the most significant
bits are used from each color channel. We use the
2 most significant bits of each color channel, for a
total of n = 64 buckets in the histogram.
For a given image I, the color histogram HI is a
compact summary of the image. A database of
images can be queried to find the most similar
image to I, and can return the image I' with the
most similar color histogram HI'. Typically color
histograms are compared using the sum of
squared differences (L2-distance) or the sum of
absolute value of differences (L1-distance). So the
most similar image to I would be the image I'
minimizing the L2-distance or L1-distance. Note
that we are assuming that weighted evenly across
different color buckets for simplicity.
Notation
M - number of pixels that an image has,
we assumed it for the ease of explanation
H(h1, h2, ..., hn) - a vector, in which each
component hj is the number of pixels of
color j in the image
n - number of distinct(discretized) color
I - an image
HI - the color histogram of image I
L2-distance
L1-distance
Description Features
Global color histogram
Uses 1 vector, H(h1, h2, ..., hn), to describes
an image.
NICE-2010
Acharya Institute of Technology, Bangalore-560090 69
hj = (the number of pixels of color j in the
image) / (total number of pixels in the
image)
Local color histogram
Divides an images into 16 equal sections
as shown in the figure below:
For each section of image, uses 1 vector,
Hk(h1, h2, ..., hn), to represent that section
of the image. So, there is 16 vectors, Hk, to
describes an image.
Default vector comparing method
In DISCOVIR, by default, we measure the
distance of 2 images as follow:
C: Texture Extraction
1. Co-occurrence
Co-occurrence matrix is a statistical method using
second order statistics to model the relationships
between pixels within the region by constructing
Spatial Gray Level Dependency (SGLD) matrices.
The Gray-level co-occurrence matrix is the two
dimensional matrix of joint probabilities Pd, r (i, j)
between pairs of pixels, separated by a distance, d,
in a given direction, r.
Notation
Pd,r(i,j) - joint probabilities between pairs
of pixels in a given direction
d - distance between pairs of pixels in a
given direction
r - a given direction
Description Features
Finding texture features from gray-level co-
occurrence matrix for texture classification are
based on thesecriteria:
Energy:
Homogeneity:
VI RESULTS
Database
A. Average RGB
Input Image
Output Images
NICE-2010
Acharya Institute of Technology, Bangalore-560090 70
B. Local Color Histogram
Input Image
Output Images
C. Global Color Histogram
Input Image
Output Images
+
D. Co-Occurance
Input Image
Output Image
CONCLUSION
Image Retrieval using Color features retrieves the
images depend on the average RGB values of an
image. It also retrieves the images depend on the
color histogram of an image. There are two types
of color histograms, Global color histograms
(GCHs) and Local color histograms (LCHs). A GCH
represents one whole image with a single color
histogram. An LCH divides an image into fixed
blocks and takes the color histogram of each of
those blocks. Texture of an image can also be
extracted and matched using Co-occurance and
edge detection methods.
Image Retrieval can be expanded with other
features of an image. Features like Shape can be
expanded. Images can be extracted using Shape
algorithms and images can be matched using
some of distance metrics.
REFERENCES
[1] John Eakins and Margaret Graham, Content-
based
Image Retrieval, JISC Technology Applications
Programme. University of Northumbria at
Newcastle.
January 1999.
[2] Christopher C. Yang, Content-based image
retrieval:
a comparison between query by example and
image
browsing map approaches , Journal of
Information
NICE-2010
Acharya Institute of Technology, Bangalore-560090 71
Science, pp. 254-267,30 (3) 2004.
[3] Rui Y. & Huang T. S., Chang S. F. Image
retrieval:
current techniques, promising directions, and
open issues.
Joumal of Visual Communication and Image
Representation,
10,39-62, 1999.
[4] Karin Kailing, Hans-Peter Kriegel and i Stefan
Schonauer, Content-Based Image Retrieval Using
Multiple Representations. Proc. 8th Int. Con. On
Knowledge-Based Intelligent Information and
Engineering Systems (KES2004), Wellington,
New Zealand, 2004, pp. 982-988
[5] Ahmed M.Ghanem, Emad M. Rasmy, and
Yasser M,Kadah, Content-Based Image Retrieval
Strategies for
Medical Image Libraries, Proc. SPIE Med. Imag.,
San Diego, Feb. 200 1 *
[6] Henning Muller, Wolfgang Muller, David McG.
Squire and Thierry Pun, Performance Evaluation
in Content-
Based Image Retrieval: Overview and Proposals.
Computing Science Center, University of Geneva,
Switzerland, 2000.
[7] A Robust Content-Based Image Retrieval
System Using Multiple Features Representations,
Mohamed A. Tahoun', Khaled A. Nagag, Taha I. El-
Arie?, Mohammed A-Megeed3
[8] Content-based Image Retrieval using Shape
and Texture Attributes Mari Partio, TAMPERE
UNIVERSITY OF TECHNOLOGY, Department of
Electrical Engineering, Institute of Signal
Processing, Nov 2002.
[9] Graph Based Segmentation in Content Based
Image Retrieval 1P.S.Suhasini, 2K. Sri Rama
Krishna and 3I.V. Murali Krishna
[10] [7 J Vishal Chitkara, Color-Based image
Retrieval Using Binary Signatures. Technical
Report TR 01-08, University of Ulberta, Canada,
May 2001.
[11] Qasim Iqbal and J. K. Aggarwal, Combining
Structure, Color, and Texture for Image Retrieval:
A performance Evaluation. 16th International
Conference on Pattern Recognition (ICPR),
Quebec City, QC, Canada, August 11-15. 2002, vol.
2, pp. 43 8-443
[12] Jiri Walder, Using 2-D wavelet analysis for
matching two images, Technical. University of
Ostrava.2000.
http://www.cg.tuwien.ac.at/studentwork/CESCG-
2000/JWalderl
[13] A Robust CBIR Approach Using Local Color
Histograms By Shengjiu Wang
[14] http://en.wikipedia.org/wiki/texture
NICE-2010
Acharya Institute of Technology, Bangalore-560090 72
CRYPTOGRAPHY USING COLORS AND ARMSTRONG NUMBERS
S. Pavithra Deepa, V. Keerthika, III B. Tech (IT)
Sri Krishna College Of Engineering and Technology,
pavithradeepa@gmail.com
_____________________________________________________________________________________________________________________________
ABSTRACT
In real world, data security plays an important
role where confidentiality, authentication,
integrity, non repudiation are given importance.
The universal technique for providing
confidentiality of transmitted data is cryptography.
This paper provides a technique to encrypt the data
using a key involving Armstrong numbers and
colors as the password. Three set of keys are used to
provide secure data transmission with the colors
acting as vital security element thereby providing
authentication.
I. INTRODUCTION:
In the present world scenario it is difficult to
transmit data from one place to another with
security. This is because hackers are becoming
more powerful nowadays. To ensure secured
data transmission there are several techniques
being followed. One among them is cryptography
which is the practice and study of hiding
information.
II. CRYPTOGRAPHY:
Cryptography, to most people, is concerned
with keeping communications private.
Encryption is the transformation of data into
some unreadable form. Its purpose is to ensure
privacy by keeping the information hidden from
anyone for whom it is not intended. Decryption is
the reverse of encryption; it is the transformation
of encrypted data back into some intelligible form.
Encryption and decryption require the use of
some secret information, usually referred to as a
key. The data to be encrypted is called as plain
text. The encrypted data obtained as a result of
encryption process is called as cipher text.
Depending on the encryption mechanism used,
the same key might be used for both encryption
and decryption, while for other mechanisms, the
keys used for encryption and decryption might be
different.
A. Types of Cryptographic Algorithms
There are several ways of classifying
cryptographic algorithms. In general they are
categorized based on the number of keys that are
employed for encryption and decryption, and
further defined by their application and use as in
[1]. The three types of algorithms are depicted as
follows
1) Secret Key Cryptography (SKC):
Uses a single key for both encryption and
decryption. The most common algorithms in use
include Data Encryption Standard (DES),
Advanced Encryption Standard (AES).
2) Public Key Cryptography (PKC):
Uses one key for encryption and another for
decryption. RSA (Rivest, Shamir, Adleman)
algorithm is an example.
3) Hash Functions:
Uses a mathematical transformation to
irreversibly "encrypt" information. MD (Message
Digest) algorithm is an example.
III. RGB COLOR FORMAT
A. RGB Color Model:
Any color is the combination of three primary
colors Red, Green and Blue in fixed quantities. A
color is stored in a computer in form of three
numbers representing the quantities of Red,
Green and Blue respectively. This representation
is called RGB representation which is used in
computers to store images in BMP, JPEG and PDF
formats. Here each pixel is represented as values
for Red, Green and Blue. Thus any color can be
uniquely represented in the three dimensional
RGB cube as values of Red, Green and Blue.
The RGB color model is an additive model in
which Red, Green and Blue are combined in
various ways to produce other colors. By using
appropriate combination of Red, Green and Blue
intensities, many colors can be represented.
Typically, 24 bits are used to store a color pixel.
This is usually apportioned with 8 bits each for
red, green and blue, giving a range of 256 possible
NICE-2010
Acharya Institute of Technology, Bangalore-560090 73
values, or intensities, for each hue. With this
system, 16777216 (256^ 3 or 2^24) discrete
combinations of hue and intensity can be
specified.
IV. PROPOSED APPROACH
A. Introduction
The existing techniques involve the use of keys
involving prime numbers and the like. As a step
further ahead let us considers a technique in
which we use Armstrong numbers and colors.
Further we also use a combination of substitution
and permutation methods to ensure data security.
We perform the substitution process by
assigning the ASCII equivalent to the characters.
Permutation process is performed by using
matrices as in [2] and Armstrong number. In this
technique the first step is to assign a unique color
for each receiver. Each color is represented with a
set of three values. For example violet red color is
represented in RGB format as (238, 58,140). The
next step is to assign a set of three key values to
each receiver.
The sender is aware of the required receiver to
whom the data has to be sent. So the receivers
unique color is used as the password. The set of
three key values are added to the original color
values and encrypted at the senders side. This
encrypted color actually acts as a password. The
actual data is encrypted using Armstrong
numbers.
At the receivers side, the receiver is aware of
his own color and other key values. The encrypted
color from the sender is decrypted by subtracting
the key values from the received set of color
values. It is then tested for a match with the color
stored at the senders database. Only when the
colors are matched the actual data can be
decrypted using Armstrong numbers. Usage of
colors as a password in this way ensures more
security to the data providing authentication. This
is because only when the colors at the sender and
receivers side match with each other the actual
data could be accessed.
Layout of the proposed technique
B. Illustration
1) Encryption
As an illustration let us assume that the data has
to be sent to a receiver (say A) who is assigned the
color raspberry (135, 38, 87). Let the key values to
be added with this color value be (-10, +5, +5). Let
the Armstrong number used for data encryption
be 153.
Step 1: (Creating password)
Password designing is a part of first phase of the
entire process.
Initially the sender knows that the required
receiver is A. So the key values are added with the
color values assigned for receiver A.
NICE-2010
Acharya Institute of Technology, Bangalore-560090 74
135 38 87
-10 5 5
-------------------------
125 43 92
Now a newly encrypted color is designed for
security check.
Step 2: (Encryption of the actual data begins
here)
Let the message to be transmitted be
CRYPTOGRAPHY. First find the ASCII equivalent
of the above characters.
C R Y P T O G R A P H Y
67 82 89 80 84 79 71 82 65 80 72
89
Step 3: Now add these numbers with the digits of
the Armstrong number as follows
67 82 89 80 84 79 71 82 65 80 72 89
(+)1 5 3 1 25 9 1 125 27 1 5 3
---------------------------------------------------
68 87 92 81 109 88 72 207 92 81 77
92
Step 4: Convert the above data into a matrix as
follows
A=
Step 5: Consider an encoding matrix...
B =
Step 6: After multiplying the two matrices (B X A)
we get
C =
The encrypted data is...
779, 3071, 13427, 890, 3598, 16082, 1383, 6075,
28431, 742, 2834, 12190
The above values represent the encrypted form of
the given message.
2) Decryption:
Decryption involves the process of getting back
the original data using decryption key. The data
given by the receiver (the color) is matched with
the data stored at the senders end. For this
process the receiver must be aware of his own
color being assigned and the ke values.
Step 1: (Authenticating the receiver)
For the receiver A (as assumed) the actual color
being assigned is Raspberry. (135, 38, 87), the key
values (set of three values) are subtracted from
the color being received to get back the original
color.
The decryption is as follows.
125 43 92 (Received data)
(-) -10 5 5 (Key values)
----------------------------------
135 38 87
(
(
(
27 125 1
9 25 1
3 5 1
(
(
(
92 92 88 92
77 207 109 87
81 72 81 68
NICE-2010
Acharya Institute of Technology, Bangalore-560090 75
The above set of values (135, 38, 87) is compared
with the data stored at the senders side. Only
when they both match the following steps could
be performed to decrypt the original data.
Step 2:
(Decryption of the original data begins here)
The inverse of the encoding matrix is
D =
Step 3: Multiply the decoding matrix with the
encrypted data (D X C) we get
Step 4: Now transform the above result as given
below
68 87 92 81 109 88 72 207 92 81 77
92
Step 5: Subtract with the digits of the Armstrong
numbers as follows
68 87 92 81 109 88 72 207 92 81 77
92
(-)1 5 3 1 25 9 1 125 27 1 5
3
------------------------------------------------------
67 82 89 80 84 79 71 82 65 80
72 89
Step 6: Obtain the characters from the above
ASCII equivalent
67 82 89 80 84 79 71 82 65 80 72
89
C R Y P T O G R A P H Y
C. Advantages
The above technique involves keys with a
minimum length of 8 bits for Armstrong numbers.
This minimum key length reduces the efforts
taken to encrypt the data. The key length can be
increased if needed, with increase in character
length. This increases the complexity thereby
providing increased security.
This technique ensures that the data transfer
can be performed with protection since it involves
two main steps. First step is to convert the
characters into another form, by adding with the
digits of the Armstrong numbers. Second step is to
encode using a matrix to form the required
encrypted data.
Tracing process becomes difficult with this
technique. This is because the Armstrong number
is used differently in each step. The key can be
hacked only if the entire steps involved in the
encoding process is known earlier.
This technique could be considered as a kind of
triple DES algorithm since we use three different
keys namely the colors, key values added with the
colors and Armstrong numbers.
Unless all the three key values along with the
entire encryption and decryption technique is
known the data cannot be obtained. So hacking
becomes difficult mainly because of the usage of
colors.
Simple encryption and decryption techniques
may just involve encoding and decoding the actual
data. But in this proposed technique the password
itself is encoded for providing more security to
the access of original data.
V. CONCLUSION
The above combination of secret key and public
key cryptography can be applied mainly in
military where data security is given more
importance. This technique provides more
security with increase in key length of the
Armstrong numbers. Thus usage of three set of
keys namely colors, additional set of key values
(
(
(
20 120 100
6 24 18
30 240 450
* ) 240 / 1 (
(
(
(
92 92 88 92
77 207 109 87
81 72 81 68
NICE-2010
Acharya Institute of Technology, Bangalore-560090 76
and Armstrong numbers in this technique ensures
that the data is transmitted securely and accessed
only by authorized people.
ACKNOWLEDGMENT
We would like to thank Mr.S.Kannimuthu,
Lecturer, Department of IT, Sri Krishna College of
Engineering and Technology for the support
provided.
REFERENCES
[1] Atul Kahate, Cryptography and Network
Security , Tata McGraw Hill Publications
[2]
http://aix1.uottawa.ca/~jkhoury/cryptography.h
tm
[3]
http://www.scribd.com/doc/29422982/Data-
Compression-and-Encoding-Using-Colors
NICE-2010
Acharya Institute of Technology, Bangalore-560090 77
DESIGN OF UHF RFID PASSIVE TAG ANTENNA FOR COMMERCIAL
APPLICATIONS USING IE3D SOFTWARE
Deshraj
Institute of technology and management, Gurgaon, MDUniversity, Rohtak
Deshrajshk580@gmail.com
_____________________________________________________________________________________________________________________________
Abstract
At the very simplest level, Radio
Frequency Identification (RFID) technologies
allow the transmission of a unique serial number
wirelessly, using radio waves. The two key parts
of the system that are needed to do this are the
RFID tag and the reader; attaching an RFID tag
to a physical object allows the object to be seen
and monitored by existing computer networks
and back-office administration systems.This
paper covers the designs and optimization of
antenna for RFID tags at UHF and microwave
frequency such design will focus on the specific
characteristics of RFID application current
development of RFID antenna that meet the
objective size reduction advance design technique
for size reduction such as mender line structure
example will be discussed. The basic need in RFID
is to miniaturize the size of tag, which contains
the antenna and an I
Keywords
Microstripe Patch Antenna, RFID,Meander Line
structure.
I. INTRODUCTION
In recent age of advanced communication
applications, greater interest is shown in printed
delectronics for low cost wireless sensors, RFID
and ambient intelligence [1]. The radio frequency
identification has become very popular in many
serves industry, distribution logistics,
manufacturing companies and goods flow system.
In this system the antenna is main device for RFID
readers. It is required to develop antennas that
are capable of maintaining high performance over
a large spectrum of frequencies .
[2].The printing process is a standard process in
packaging industry and electronic labels in the
form of In recent age of advanced communication
application greater interest is shown in printed
electronics for low cost wireless sensors, RFID
and ambient intelligence.
[1]. the radio frequency printed RFID tags is very
much desired and much efforts are undertaken to
reach expected standards for patch antenna and
tags. This is basically due to multiple services
having large amount of information, which needs
very high bandwidth. This technological trend has
brought changes is antenna design and one of the
design is micro strip (patch) antennas [2].
Presently micro strip patch antennas are popular
in many wireless applications due to their
advantages when compared with conventional
microwave antennas. The patch antenna
geometry is light weight, low volume and thin
profile configurations. This offers low cost and
amenable to mass production. The patch antennas
can easily be fabricated for circular or dual
polarization and very easily fabricated on
microwave IC with feed and matching network.
These can be fabricated using modern day
printing circuit board technology, compatible with
MMICs, and have the ability to conform to planar
and non-planar surfaces [3]. Due to having
specific features all of above characteristics it is
very useful to use the microstrip antennas in the
radio frequency identification (RFID) applications.
One of the main characteristics of patch antennas
is their inherently narrowband performance due
to its resonant nature, which helps to get a sharp
resonant frequency [4].
II. PROPOSED GEOMETRIES
For radio communication system
applications, Omni-directional antenna radiation
is preferred when coverage area is direction free
and need signal in a direction as the case of RFID
is. Antenna performance at UHF and microwave
frequency are dependent on the substrate i.e. and
electromagnetic properties(conductivity,
permittivity and permeability). It also depends on
the conductive surface geometry. In this paper our
main concern is to change the conductive surface
geometry dimensions and then visualize the effect
of frequency. In the proposed design we
considered Meander line structure for getting
better performance with compact design having
greaterefficiency. An Omni-directional pattern can
be accomplished by a small loop antenna whose
circumference may be less than one-tenth of
wavelength due to its inherent property but due
to small input resistance it is very difficult to
realize. This has led to array antennas with
complex feed systems.
NICE-2010
Acharya Institute of Technology, Bangalore-560090 78
Fig.1(a). Proposed Geometry (3.5 steps)
Fig. 1(b). Full Step Dimension
Fig. 1(c). half Step Dimension
In figure 1 the probe fed Meander Line
RFID antenna with meandered structure of a
finite width strip is shown. Due to this design we
get resonance wit a much more compact structure
because we have a large electrical length in a
small planar structure which waves will travel
and results the radiated field of the antenna [5].
For proposed designs with variations in points we
are considering the following parameters:
Height of substrate 0.05mm (2mil )Dielectric
constant of substrate (PET) 3.50 Loss tangent
0.017 In the proposed designs of meander line
antenna all the dimension are indicated in mm.
In figure1(a) we have shown the proposed
compact geometry of meander line structure with
3.5 steps in centre and fig 1(b) is showing full step
dimension for proposed design and in figure 1(c)
the half step dimension is shown. The proposed
design with meander structure has been
fabricated and tested, which has given response in
accordance to available theoretical standards and
conventions. The meander line design has further
reduced the size of antenna for readers.
III. RESULTS
For proposed design given in previous
figures the frequency versus return loss graphs
are shown in this section with various changes
and their outcome. In figure 2, graph has been
plotted between frequency and return loss for 3.5
step dimension. It is clear from the graph that it
has resonated 700 MHz. After achieving the graph
for 3.5 step dimension the variation in steps has
been done and analysed for changes in resonance.
Fig 2 Frequency vs. Return loss [S11]
To analyse the effect of step changes in
proposed meander line design the step changes in
geometry have been shown in figures 3(a), 4(a),
5(a) 6(a) and 7(a) for geometry steps 3.0, 2.5, 2.0,
1.5 and 1.0 respectively. While the effect of
geometry step changes have been shown in
frequency graphs in figures 3(b), 4(b), 5(b), 6(b)
and 7(b). the specific meander line design with
return loss graphs are also shown in fig. 8 and 9.
The Return Loss graph we have clearly shown for
865 and 915 MHz Which is desired frequency of
RFID tag. The effect of changes in geometry steps
and their corresponding frequency of resonance is
also shown in table no. 1 clearly. This will help in
proposing new designs with shift in frequency for
some specific features.
Table No. 1: meander line step dependency
Number of step Frequency(MHz)
3.5 700
3.0 800
NICE-2010
Acharya Institute of Technology, Bangalore-560090 79
2.5 900
2.0 1000
1.5 1150
Fig. No. 3(a) 3.0 step geometry
Fig. No. 3(b) 3.0 Freq vs Return loss [S11]
Fig. No. 4(a) 2.5 step geometry
Fig. No. 4(b) 2.5 Freq vs Return loss [S11]
Fig. 5(a) 2.0 step geometry
Fig. No. 5(b) 2.0 Freq vs Return loss [S11]
Fig. No. 6(a) 1.5 step geometry
Fig. No. 6(b) 1.5 Freq vs Return loss [S11]
NICE-2010
Acharya Institute of Technology, Bangalore-560090 80
Fig. 7(a) Geometry for Frequency 865 MHz
Fig. No. 7(b) Freq vs Return loss [S11]
Fig. 8(a). Geometry for Frequency 915 MHz
Fig. 8(b). Frequency vs. Return loss [S11]
IV. CONCLUSION
Radio frequency identification is a rapidly
developing technology for automatic
identification of object. In this paper, we
presented an overview of antenna design for
passive UHF RFID tags antenna. In this paper we
have shown effect of geometry step change
affecting the resonance frequency of antenna. The
proposed design is simple structure, and can be
constructed with a very low cost. In addition, good
antenna gain and radiation patterns have also
been obtained. We presented a meander line
antenna design for standard frequencies of RFID
with analysis results after fabrication. The design
was simulated in IE3D v.14.1 and fabricated. The
results have justified meander line applications
with optimal simulation results.
V. REFRENCES
[1] H-E. Nilsson, J. Siden, T. Olsson, P. Jonsson and
A. Koptioug, Evaluation of Printed Patch antenna
for robust microwave RFID tags, IET Microw.
Antennas Propag., 2007, 1, (3), pp. 776781
[2] C. Balanis, Antenna Theory Analysis and Design,
and 2
nd
ed: John Wiley & Sons, Inc., 1997.
[3] J.R.James and P.S.Hall,Handbook of Microstrip
Antennas Peter Peregrinus Ltd., IEE, 1989.
[4] Smail Tedjini, Tan-Phu Vuong and Vincent
Beroulle Antennas for RFID tags in joint sOc-
EUSAI conference, pp.19-22, Oct.2005.
[5] K.V. Seshagiri Rao, Pavel V. Nikitin and Sander
F. Lam Antenna Design for UHF RFID Tags: A
Review and a Practical Applicationin IEEE Trans.
pp. 3870-3876 Vol.53, No.12,Dec.2005
NICE-2010
Acharya Institute of Technology, Bangalore-560090 81
EXACT SENSITIVE KNOWLEDGE HIDING THROUGH DATABASE EXTENSION
Muthukarthick B, Kesavan R (III Year)
Department of Information Technology
Velammal College of Engineering & Technology, Madurai
bmuthukarthik@gmail.com, Mob - +919994276153
________________________________________________________________________________________________________________________
ABSTRACT
Sharing data among organizations often
leads to mutual benefit. Recent technology in data
mining has enabled efficient extraction of
knowledge from large databases. This, however,
increases risks of disclosing the sensitive knowledge
when the database is released to other parties. To
address this privacy issue, one may sanitize the
original database so that the sensitive knowledge is
hidden. Sensitive knowledge hiding in large
transactional databases is one of the major goals of
privacy preserving data mining. A novel approach
that performs sensitive frequent item set hiding by
applying an extension to the original database was
proposed. The extended portion of data set
contains transactions that lower the importance of
the sensitive patterns, while minimally affecting the
importance of the non-sensitive ones. We present
the border revision to identify the revised border
for the sanitized database and then we compute the
minimal size for the extension. The hiding process
involves the construction of a Constraints
Satisfaction Problem, by using item sets of revised
borders and its solution through Binary Integer
Programming.
KEYWORDS
Privacy preserving data mining, knowledge
hiding, and association rule mining, binary integer
programming.
I. INTRODUCTION
A subfield of privacy preserving data
mining is knowledge hiding. The paper presents
a novel approach that strategically performs
sensitive frequent itemset hiding based on a new
notion of hybrid database generation. This
approach broadens the regular process of data
sanitization by applying an extension to the
original database instead of either modifying
existing transactions, or rebuilding the dataset
from scratch. The extended portion of the dataset
contains a set of carefully crafted transactions
that achieve to lower the importance of the
sensitive patterns to a degree that they become
uninteresting from the perspective of the data
mining algorithm, while minimally affecting the
importance of the nonsensitive ones. The hiding
process is guided by the need to maximize the
data utility of the sanitized database by
introducing the least possible amount of side-
effects, such as (i) the hiding of non-sensitive
patterns, or (ii) the production of frequent
patterns that were not existent in the initial
dataset. The released database, which consists of
the initial part (original database) and the
extended part (database extension), can
guarantee the protection of the sensitive
knowledge, when mined at the same or higher
support as the one used in the original database.
The approach introduced in this paper is exact in
nature. On the contrary, when an exact solution is
impossible, the algorithm identifies an
approximate solution that is close to the optimal
one. To accomplish the hiding task, the proposed
approach administers the sanitization part by
formulating a Constraint Satisfaction Problem
(CSP) and by solving it through Binary Integer
Programming (BIP).
II. KNOWLEDGE HIDING FORMULATION
This section provides the necessary background
regarding sensitive itemset hiding and sets out
the problem at hand, as well as the proposed
hiding methodology.
A. Frequent Itemset:
Let I = {i1, i2,.. iM} be a finite set of literals, called
items, where M denotes the cardinality of the set.
Any subset I I is an item set over I. A transaction
T over I is a pair T = (tid, I), where I is the item set
and tid is a unique identifier, used to distinguish
among transactions that correspond to the same
item set. A transaction database D = {T1, T2 ...TN}
over I is an
N x M table consisting of N transactions over I
carrying different identifiers, where entry Tnm = 1
if and only if the mth item (m [1, M]) appears in
the nth transaction (n [1, N]). Otherwise, Tnm =
0. A transaction T = (tid, J) supports an item set I
over I, if I J. Let S be a set of items; notation p(S)
denotes the power set of S, which is the set of all
subsets of S.
Given an item set I over I in D, sup (I, D)
denotes the number of transactions T D that
NICE-2010
Acharya Institute of Technology, Bangalore-560090 82
support I and freq (I, D) denotes the fraction of
transactions in D that support I. An item set I is
called large or frequent in database D if and only if
its frequency in D is at least equal to a minimum
threshold mfreq. A hybrid of Apriori and FP-Tree
algorithms are proposed to be used to find the
frequent item set.
A hybrid of Apriori and FP-Tree
algorithms are proposed to be used to find an
optimized set of frequent item set.
1) The Apriori Algorithm - Finding Frequent
Itemset Using Candidate Generation:
Apriori is an influential algorithm for
mining frequent itemset for Boolean
association rules. Apriori employs an iterative
approach known as a level-wise search, where
k-itemset are used to explore (K+1) itemset.
First, the set of frequent 1-itemsets is found.
This set is denoted L1. L1 is used to find L2, the
set of frequent 2-itemsets , which is used to
find L3 , and so on, until no more frequent k-
itemsets can be found. The finding of each Lk
requires one full scan of the database.
To improve the efficiency of the level-wise
generation of frequent itemsets, an important
property called the Apriori property, i.e., all
nonempty subsets of a frequent itemset must also
be frequent, is used to reduce the search space. A
two step process join and prune actions are used
to find Lk from Lk-1.
2) FP-Growth Algorithm-Mining Frequent
patterns without candidate generation:
Frequent pattern growth or simply FP-growth
adopts a divide-and conquer strategy as follows:
compress the database representing frequent
items into a frequent pattern tree, or FP tree, but
retain the itemset association information and
then divide such a compressed database into a set
of conditional databases, each associate with one
frequent item, and mine each such database
separately.
Major steps to mine FP-tree
1) Construct conditional pattern base for
each node in the FP-tree
2) Construct conditional FP-tree from each
conditional pattern-base
3) Recursively mine conditional FP-trees and
grow frequent patterns obtained so far
If the conditional FP-tree contains
a single path, simply enumerate
all the patterns
B .Hiding methodology
To properly introduce the hiding methodology,
one needs to consider the existence of three
databases, all depicted in binary format. They are
defined as follows:
Database Do, is the original transaction
database that, when mined at a certain
support threshold msup, and leads to the
disclosure of some sensitive knowledge in the
form of sensitive frequent patterns. This
sensitive knowledge needs to be protected.
Database Dx ,is a minimal extension of Do that
is created by the hiding algorithm during the
sanitization process, in order to facilitate
knowledge hiding.
Database D, is the union of database Do and
the applied extension Dx and corresponds to
the sanitized outcome that can be safely
released.
TABLE 1
Sample: Sanitized Database D as a Mixture of the
Original Database DO and the Applied Extension
DX
Table 2
Sample: Frequent Item Sets for DO and DX at msup
= 3 (for table 1)
NICE-2010
Acharya Institute of Technology, Bangalore-560090 83
III. HYBRID SOLUTION METHODOLOGY
A. Computation of size of database extension
Database DO is extended by DX to
construct database D. An initial and very
important step in the hiding process is the
computation of the size of DX. A lower bound on
this value can be established based on the
sensitive item set in S, which has the highest
support. The rationale here is given as follows: by
identifying the sensitive item set with the highest
support, one can safely decide upon the minimum
number of transactions that must not support this
item set in DX, so that it becomes infrequent in D.
Lower bound Q is
1
) , sup(
+
(
= N
mfreq
D I
Q
O N
.1
Equation (1) provides the absolute minimum
number of transactions that need to be added in
DX, to allow for the proper hiding of the sensitive
item sets of DO. However, this
lower bound can, under certain circumstances, be
insufficient to allow for the identification of an
exact solution, even if one exists. To circumvent
this problem, one needs to expand the size Q of DX
as determined by (1), by a certain number of
transactions. A threshold, called safety margin
(denoted hereon as SM), is incorporated for this
purpose. Safety margins can be either predefined
or be computed dynamically, based on particular
properties of database DO and / or other
parameters regarding the hiding process.
B. Exact and Ideal Solutions
Definition 1 (feasible/exact/approximate
solution).
A solution to the hiding of the sensitive
knowledge in DO is considered as feasible if it
achieves to hide the sensitive patterns. Any
feasible solution, introducing no side effects in the
hiding process, is called exact. Finally, any non
exact feasible solution is called approximate.
Definition 2 (database quality).
Given the sanitized database D, its original
version DO, and the produced extension DX , the
quality of database D is measured both in the size
of DX and in the number of binary variables set to
1 in the transactions of DX (i.e., the distance
metric). In both cases, lower values correspond to
better solutions.
Definition 3 (ideal solution).
A solution to the hiding of the sensitive
item sets is considered as ideal if it has the
minimum distance among all the existing exact
solutions and is obtained through the minimum
expansion of DX. In that sense, ideal is a solution
that is both minimal (with respect to distance and
size of extension) and exact.
C. Border Revision
The rationale behind this process is that
hiding of a set of item sets corresponds to a
movement of the original borderline in the lattice
that separates the frequent item sets from their
infrequent counterparts , such that the sensitive
item sets lie below the revised borderline. There
are four possible scenarios involving the status of
each item set I prior and after the application of
border revision:
C1 :
Item set I was frequent in DO and remains
frequent in D.
C2 :
Item set I was infrequent in DO and is infrequent
in D.
C3 :
NICE-2010
Acharya Institute of Technology, Bangalore-560090 84
Item set I was frequent in DO and became
infrequent in D.
C4 :
Item set I was infrequent in DO and became
frequent in D.
Since the borders are revised to
accommodate for an exact solution, the revised
hyper plane is designed to be ideal in the sense
that it excludes only the sensitive item sets and
their supersets from the set of frequent patterns
in D, leaving the rest of the item sets in their
previous status as in database DO.
The first step in the hiding methodology
rests on the identification of the revised borders
for D. The hiding algorithm relies on both the
revised positive and the negative borders,
denoted as Bd
+
(F
1
D) and Bd
(F
1
D), respectively.
After identifying the new (ideal) borders, the
hiding process has to perform all the required
minimal adjustments of the transactions in Dx to
enforce the existence of the new borderline in the
result database.
Fig. 1 An sample item set lattice demonstration (a)
the original border and the sensitive item sets, and
(b) the revised border for Table1
D. Problem Size Reduction
To enforce the computed revised border and
identify the exact hiding solution, a mechanism is
needed to regulate the status (frequent versus
infrequent) of all the item sets in D. Let C be the
minimal set of border item sets used to regulate
the values of the various uqm variables in DX.
Moreover, suppose that I C is an item set, whose
behavior we want to regulate in D. Then, item set I
will be frequent in D if and only if sup (I, Do) +
Sup(I, Dx) m freq x (N+Q), or equivalently if
Sup (I,DO) +
=
Q
q I i
M
1
uqm mfreq x (N + Q) (3)
equivalently if
sup (I,DO) +
=
Q
q I i
M
1
uqm < mfreq x (N + Q) ...(4)
Inequality (3) corresponds to the minimum
number of times that an item set I has to appear in
the extension DX to remain frequent in D. On the
other hand, (4) provides the maximum number of
times that an item set I has to appear in DX to be
infrequent in database D. To identify an exact
solution to the hiding problem, every possible
item set in P, according to its position in the
latticewith respect to the revised bordermust
satisfy either (3) or (4). However, the complexity
of solving the entire system of the 2
M
_ 1
inequalities is well known to be NP-hard .
Therefore, one should restrict the problem to
capture only a small subset of these inequalities,
thus leading to a problem size that is
computationally manageable. The proposed
problem formulation achieves this by reducing
the number of the participating inequalities that
need to be satisfied. Even more, by carefully
selecting the item sets of set C, the hiding
algorithm ensures that the exact same solution to
the one of solving the entire system of inequalities
is attained. This is accomplished by exploiting
cover relations existing among the item sets in the
lattice.
Set C is chosen appropriately to consist of
all the item sets of the revised border. The
proposed hiding algorithm is capable of ensuring
that if (3) and (4) are satisfied for all the item sets
in C, then the produced solution is exact and is
identical to the solution involving the whole
system of the 2
M
1 inequalities. Cover relations
governing the various item sets in the lattice of DO
ensure that the formulated set of item sets C has
an identical solution to the one of solving the
system of all 2
M
1 inequalities for D.
NICE-2010
Acharya Institute of Technology, Bangalore-560090 85
The cover relations that exist between the
item sets of Bd
+
(F`D) and those of F`D. In the same
manner, the item sets of Bd
(F`D)
Based on (7), the item sets of the revised borders
Bd
+
(F`D) and Bd