Digital Signal Processing 68 (2017) 112–126

Contents lists available at ScienceDirect

Digital Signal Processing

www.elsevier.com/locate/dsp

Anisotropic diffusion based denoising on X-radiography images to

detect weld defects
Muthukumaran Malarvel a , Gopalakrishnan Sethumadhavan a,∗ ,
Purna Chandra Rao Bhagi b , Soumitra Kar c , Thangavel Saravanan b , Arunmuthu Krishnan b
a
School of Computing, SASTRA University, Thanjavur, Tamil Nadu, India
b
Indira Gandhi Centre for Atomic Research, Kalpakkam, Tamil Nadu, India
c
Bhabha Atomic Research Centre, Mumbai, Maharashtra, India

a r t i c l e i n f o a b s t r a c t

Article history: This paper proposes a machine vision scheme for denoising, feature space gradient preserving, and
Available online 3 June 2017 detecting weld defects in noisy weld X-radiography images; particularly, for the images that are in low-
contrast and contain noises. The detection of small weld defects present on noisy image is extremely
Keywords:
difficult in non-destructive testing through machine vision. The presence of high gradient magnitude and
Anisotropic diffusion
the low intensity in the feature space of a noisy image are the main characteristics of weld defects.
Image denoising
Weld defects These characteristics can be considered to refine and obtain noise-free images for detection of weld
Segmentation defects. This study proposes a modified anisotropic diffusion model, which considers a local probability
X-radiography image value of gray-level and an adaptive threshold parameter in diffusion coefficient function to adjust the
implication of low edge gradient of the feature space from the noisy image. Furthermore, an entropy
based stopping criterion has been introduced to terminate the diffusion process. This proposed model is
compared with the existing models, and its performance is evaluated through Mean Square Error (MSE),
Signal-to-Noise Ratio (SNR), Peak Signal-to-Noise Ratio (PSNR), Entropy (E) and Mean Structural Similarity
(MSSIM) measures. Experimental results confirm the reliability of the proposed model.
© 2017 Elsevier Inc. All rights reserved.

1. Introduction
Ultrasonic, Eddy current, and X-radiography inspection tech-
niques are commonly used in non-destructive evaluation industry.
Digital X-radiography has widely been used in Non-Destructive
Testing (NDT) and particularly in weld defect detection [1]. The
detection of defects depends on the quality of the digitized image,
which is subjected to certain factors, such as noise, low contrast,
defects of different dimensions, and indiscernible defects in the
image background. In experimental sciences, noise refers to any
random fluctuations of data that hinders perception of an expected
signal. The noise is a summation of unwanted or disturbing energy
from natural or man-made sources, and is typically distinguished
from interference or distortion. Sources such as automobiles, air-
craft, ignition electric motors and switching gear, high voltage
wires and fluorescent lamps cause industrial noise, produced by
the discharge present in all these operations. Gaussian noise is sta-
tistical noise having a Probability Density Function (PDF) equal to
* Corresponding author.
E-mail address: sgk@mca.sastra.edu (G. Sethumadhavan).
that of the normal distribution, i.e., the values that the noise can
take on are Gaussian-distributed. In digital images, including those
in X-radiography, it is common practice amongst the researchers to
model the present noise as Gaussian and apply suitable denoising
filters on the image for its quality improvement. Gaussian noise is
also called electronic noise because it arises during image digiti-
zation process, which is independent of original pixel values and
interrupts the gray values during digitization. Moreover, the noisy
digital weld X-radiography image reduces the ability to identify
tiny weld defects, and it is further worsened if the image con-
taminated by noises. However, the features present in images are
degraded or contaminated by the unknown Gaussian noise levels.
Therefore, it is a challenging task to remove such noises with-
out disturbing the image features during noise removal process.
To devise a suitable model to remove the unknown Gaussian noise
present in the digital weld X-radiography images during denoising
and smoothing process therefore turns out to be important.
Incorrect feature extraction leads to detection of inaccurate de-
fects. Moreover, preserving the features present in the Region of
Interest (ROI) is considered as severe problem and it is a pre-
requisite for image interpretation and analysis during denoising
and smoothing processes. In this regard, several denoising and

http://dx.doi.org/10.1016/j.dsp.2017.05.014
1051-2004/© 2017 Elsevier Inc. All rights reserved.
 M. Malarvel et al. / Digital Signal Processing 68 (2017) 112–126
smoothing techniques of edge-preserving have been approached to
remove noises and revamp the features of an image without dis-
torting its region and region edges in the image.
Among the denoising and smoothing methods, nonlinear meth-
ods have been found to deliver more adequate results in com-
parison to linear methods. Several nonlinear approaches using
Anisotropic Diffusion Filter (ADF), fourth order Partial Differential
Equation (PDE), Non-Local Means (NLM), combination of domain
and range filters, nonlinear Gaussian filters and bilateral filters and
their extensions to the trilateral filters [2,3], and are widely known.
In this paper, an anisotropic diffusion based scheme is proposed to
solve the problem of weld defect detection in the noisy images.
Perona and Malik were the first to apply the anisotropic diffusion
equation for image smoothing and edge detection. This classical
method is known as P–M Model [4], which has developed for
edge detection, enhancement of image, smoothing the image, im-
age restoration, segmentation of features in the image. In [5,6], two
approaches have been proposed followed by P–M model, which
have motivated to introduce many approaches till date. Catte et al.
[5] proposed a scheme that carries out pre-denoising with a Gaus-
sian filter before each iteration of diffusion function. This scheme
can efficiently reduce noise; however, the selection of a suitable
value for σ parameter of Gaussian filter is still a problem. Rudin et
al. [6] proposed nonlinear total variation (TV) method. This scheme
provides a better effect on staircases using gradients alone.
In recent years, several modified anisotropic diffusion (AD)
models with different schemes have been approached for digital
image processing. The modified AD models can be classified into
two categories based on denoising process and edge preserving.
Regarding the denoising methods that are using local information
for diffusion process, Chao and Tsai et al. proposed several ap-
proaches [7–12] based on anisotropic diffusion for noise removal
and defect detection in sputtered glass, astronomical images, Thin
Film Transistor images, solar wafer, and surface images. Tsai and
Chao [7] formulated an anisotropic diffusion (TCAD) method for
defect detection in sputtered glass substrates with inhomogeneous
textures. This method employs a non-negative reducing function
with a strengthening gradient threshold as the diffusion coeffi-
cient to adjust adaptively the importance of edge gradients. Chao
and Tsai [8] further proposed an improved anisotropic diffusion
method, known as CTAD. In order to eliminate the noisy stars from
a nebula image, this method incorporated the local variance infor-
mation of gray-levels in the diffusion coefficient function. Another
version of diffusion model [9] called CTAD-TFT has successfully
sharpened and smoothened low-contrast surface images, particu-
larly, in glass substrates used in Thin Film Transistor-Liquid Crystal
Displays. Weighted sharpening diffusion function was employed in
the classical P–M model for smoothing process.
In [10], micro-cracks in polycrystalline solar wafer surface im-
ages involving heterogeneous texture patterns were detected using
anisotropic diffusion. The normalized gray-level of an original im-
age was used as diffusion coefficient function for smoothing pro-
cess. In this method, two post processing steps, binary threshold-
ing and morphological operations, were executed to segment the
defects after the smoothing of image. Generalized diffusion coeffi-
cient function [11] in anisotropic diffusion was proposed to detect
defects in low-contrast surface images that can adaptably adjust
the curve of the diffusion coefficient function. This method ap-
plies a stochastic evolutionary computation algorithm and particle
swarm optimization (PSO) to determine the parameter values au-
tomatically. A modified anisotropic diffusion technique named CTD
for an edge-preserving smoothing was proposed [12]. This method
preserves edges besides fine details during noise removing in the
diffusion process. It employs both local gradient and normalized
variance of gray-level in diffusion coefficient function with adap-
tive selection of threshold parameter.
113
Tian et al. [13] proposed another modified version of P–M
model, a Difference Eigenvalue Perona Malik (DEPM) model, which
preserves fine features during noise removal processes. This model
was adapted and an edge indicator called difference eigenvalue
to measure image gradient magnitude in the diffusion process
was introduced. Surya Prasath and Vorotnikov [14] proposed a
scheme for image denoising and restoration called weighted and
well-balanced anisotropic diffusion. In order to remove noise with-
out loss of fine details, a new variant of the flow is considered
using weight parameter in the divergence diffusion process.
Another category includes edge-preserving approaches that are
based on changing the diffusion function or introducing new oper-
ations [15–23]. Barcelos et al. [15] approached the Well-Balanced
Flow (WBF) based on Total Variation (TV), which is used in the
divergence process of the P–M model. The Gaussian kernel was uti-
lized as pre-smoothing to avoid noise oscillations. Mean curvature
flow equation and Nordstrom’s biased PDE were employed in WBF
model. Coupled anisotropic diffusion [16] scheme was introduced
using the PDE systems consisting of AD equations. This scheme
will smooth regions during the diffusion function, where the gra-
dient is trivial, but will preserve it where the gradient is large on
a noisy image. A kernel anisotropic diffusion (KAD) [17] method
utilizes a kernalized gradient operator in the diffusion function for
noise removal and detection of edge in a noisy image. To terminate
the diffusion procedure automatically, the KAD introduces adap-
tive diffusion threshold estimation using Median Absolute Devia-
tion (MAD) of the gradient magnitude and employs Mean Absolute
Error (MAE). Dai Fang et al. [18] proposed a nonlinear diffusion
model for smoothing and sharpening of image based on the Mean
Curvature Motion (MCM) and the inverse heat diffusion equations.
The staircase effect in the image has been considered in the Ramp
Preserving Perona–Malik model (RPPM) [19]. The RPPM model in-
troduces isotropic and anisotropic diffusion properties to preserve
staircase effect and edge region respectively. Guo et al. [20] pro-
posed an Adaptive Peronal–Malik (APM) model for image denoising
based on variable exponent. The APM roughly segments the image
into two regions using the edge indicator. Among these regions,
APM remove the noise and preserve the edge accordingly. Wang
et al. [21] proposed a Modified Perona–Malik (MPM) model based
on directional Laplacian. In order to remove noise at the same time
preserving edges, a weighted Laplacian was incorporated where di-
rectional Laplacian diffuses image along the edge direction of the
original image. Maiseli and Gao [22] proposed a diffusion-driven
iterative method to extract edges from noisy images. This method
utilized anisotropic diffusion to suppress noise and introduced an
edge-sensitive detector to highlight the edges of the feature. Xu et
al. [23] have proposed a method based on semi-adaptive threshold
in anisotropic diffusion process. Gaussian filter used to distinguish
corrupted pixels and noise-free pixels as a pre-denoising process
then in the post-processing, a semi-adaptive threshold function in
diffusion coefficient is used.
On weld X-radiography image processing domain, in particu-
lar denoising and smoothing process, very fewer researchers have
applied AD based approach in detecting weld defects [24–26].
Mhamed et al. [24] proposed a modified AD model for automatic
defect detection in X-radiography images. This model initially con-
volutes the gradient image using median filter. Then the edge
sharpening function was added to the diffusion coefficient function
for weld defect detection. Halim et al. [25] proposed a model based
on PDE for weld defect detection on digital radiography image. This
method utilized P–M model as a denoising process for smooth-
ing, and Chan–Vese method for segmentation and defect detection
process. Dang et al. [26] considered the P–M anisotropic diffusion
filter as an image smoothing besides preserving the edges of the
defects. In this approach, linear weighting and fuzzy enhancement
algorithms are involved in segmentation of weld defect.
114 M. Malarvel et al. / Digital Signal Processing 68 (2017) 112–126

Although the existing AD based image-denoising techniques

adaptively regulate the diffusion process depending on the di-
verse information in diffusion coefficient function, they still have
some drawbacks. Furthermore, the existing methods randomly as-
sign a constant threshold parameter in diffusion coefficient func-
tion, which makes the smoothing process ineffective; and require
more number of iterations to complete the smoothing process.
In this paper, we propose a modified anisotropic diffusion model
that improves the denoising by feature preserving smoothing to
detect the weld defects in noisy X-radiography images. This modi-
fied P–M model considers local gray-level probability value of the
3 × 3 window size, and assigns the entropy of the image to the
threshold parameter, which is incorporated with diffusion coeffi-
cient function. Furthermore, the entropy based stopping criterion
Fig. 1. Graph of the diffusion coefficient function: g (∇ I ) = 1/[1 + (|∇ I |/k)2 ].
in the diffusion process is also introduced to minimize the smooth-
ing iterations and terminating the diffusion process.

2. Anisotropic diffusion models

2.1. Classical Perona–Malik model (P–M)

Perona and Malik [4] proposed anisotropic diffusion method

based on heat conduction equation, which can be given by
⎧
⎨ ∂ I t (x, y ) = divc (x, y ) · ∇ I (x, y )
t t
∂t (1)
⎩
I t =0 = I o

where I o is the original image at iteration t = 0, div is the diver-

gence operator, ∇ I t (x, y ) is the gradient of the image and ct (x, y )
Fig. 2. Graph of the flux function: φ(∇ I ) = 1/[1 + (|∇ I |/k)2 ] · ∇ I .
is the diffusion coefficient.
The anisotropic diffusion in Eq. (1) can be discretely determined
by using the four nearest neighbor pixels, N 4 (x, y ) and the Lapla- In P–M model, the diffusion coefficient g (∇ I ) is fixed adaptively
cian operator [4]. It is given by in every iteration that differentiates the edges present in the im-
age by reducing the diffusion along the edge directions. Since the
1  (i )
4
(i ) diffusion process in P–M model is unstable, and the artifacts are
I t +1 (x, y ) = I t (x, y ) + ct (x, y ) · ∇ I t (x, y ) (2)
4 still occur in the image [23].
i =1
The selection of threshold k is a challenging task and it is es-
(i )
where ∇ I t (x, y ), for all i = 1 to 4, denote the gradients of 4 sential to fine-tuned; if the higher k value in the diffusion process
neighbors in N 4 (x, y ). Hence, delivers over smooth the image and thereby the feature region is
also more blurred. In contrast, if the lower k value in the diffusion
(1 ) process delivers under smooth and thereby the noises will not be
∇ I t (x, y ) = I t (x, y − 1) − I t (x, y );
removed appropriately. However, the feature region is still contam-
(2 )
∇ I t (x, y ) = I t (x, y + 1) − I t (x, y ); inated by noise. The selection of desired k will be visualized and
(3 ) quantified through the flux function.
∇ I t (x, y ) = I t (x + 1, y ) − I t (x, y );
Let φ(∇ I ) be the flux function [4] defined by
(4 )
∇ I t (x, y ) = I t (x − 1, y ) − I t (x, y ).
(i ) (i ) φ(∇ I ) = g (∇ I ) · ∇ I (5)
ct (x, y ) denotes the diffusion coefficient related with ∇ I t (x, y ),
(i )
and is expressible as a function of the gradient ∇ I t (x, y ) in the The diffusion coefficient function and the flux function of
P–M model: Eq. (4) and Eq. (5) are demonstrated in Fig. 1 and Fig. 2, respec-
tively. In Fig. 1, it can be seen that the gradient magnitude |∇ I |


(i )
ct (x, y ) = g ∇ I t (x, y ) .
(i ) monotonically decreasing and converging near to zero when |∇ I |
greater than 5k. Hence, the diffusion stops earlier as |∇ I | > 5k. In
The following two diffusion coefficient functions, which are widely Fig. 2, the higher smoothness achieved when |∇ I | = k. Therefore,
used in the diffusion process, are given by Perona–Malik as Eqs. (3) the P–M model can efficiently smooth intra-regions, which is de-
and (4) below: fect free background region in an image.
 2 
|∇ I |
g (∇ I ) = exp − and (3) 2.2. Modified Perona–Malik models
k
1 2.2.1. TCAD model
g (∇ I ) = |∇ I | 2
(4)
1+( ) Tsai and Chao (TCAD) [7] have proposed a model for detection
k
of defects present in sputtered surfaces with non-homogeneous
where k is a fixed threshold value which regulates the denoising textures using anisotropic diffusion. An annealing cubic-root func-
level. tion is adapted in this model. The TCAD model presented as:
 M. Malarvel et al. / Digital Signal Processing 68 (2017) 112–126 115

(i ) 1
g (∇ I t (x, y ) = (i )
(6)
|∇ I (x, y )|
1 + [ k(0t)t −(1/3) ]2

where k(0) is the average gradient magnitude of the entire image

and is calculated as
 
4 M −1 N
 −1
1 (i )
k(0) = ∇ I t (x, y ) (7)
4M N
i =1 x=0 y =0

where M × N is the spatial resolution of the image.

2.2.2. CTAD model

The CTAD model [8] was proposed to handle the image restora-
tion in astronomical images. The local variance of the gray level Fig. 3. 3D plot of the diffusion coefficient function g (∇ I t (x, y ), pt
(i ) S xy
(x, y )) with
incorporated in this diffusion model. In this model, the modified (i )
variables ∇ I t (x, y ) and pt
S xy
(x, y ).
diffusion coefficient function is given as follows:

(i )  1 3. Proposed anisotropic diffusion model
g ∇ I t (x, y ), σ 2 = (i )
(8)
|∇ I (x, y )|
1 + [ kt ·σ 2 ]2 To determine the diffusion coefficient functions in TCAD, CTAD,
where σ is the local variance of gray levels in a 3 × 3 neighbor-
2 CTAD-TFT and CTD models, the annealing cubic-root, variance
hood window. The variance is defined as of the gray-level, sharpening, and normalized variance of the
gray-level functions are used respectively, instead direct use of
1  

1 1
2 anisotropic diffusion. In this paper, we propose a revised AD model
σt2 (x, y ) = I t (x + i , y + j ) − I t (x, y ) (9) that employs both probability value of local gray-level and gradi-
9
i =−1 j =−1 ent for denoising and smoothing the noisy image adaptively. The
proposed method not only preserves edges, but also retains the
where I t (x, y ) is the mean of the gray levels in the 3 × 3 neigh-
fine details of the feature during smoothing process. The proposed
borhood window at iteration t.
model is explained in the following Sections 3.1 and 3.2.
2.2.3. CTAD-TFT
The CTAD-TFT model [9] combines both the smoothing and 3.1. Reduction of noise
sharpening processes in a single function and the same has been
exhibited on surface images of glass substrates used in TFT-LCDs. A pixel at coordinate (x, y ) and at iteration t, the gray-level
In order to improve the subtle defects in the image, this model in- probability is computed from its 3 × 3 neighborhood region, S xy
corporates the weighted sharpening process in the diffusion func- as:
tion. The sharpening diffusion coefficient function ν (∇ I ) is defined S xy I t (x, y )
as pt (x, y ) = , (14)
I t (x, y )

ν (∇ I ) = α · 1 − g (∇ I ) (10) where I t (x, y ) denotes the sum of gray-levels in the 3 × 3 neigh-
where α is the weight of sharpening coefficient function with 0 ≤ borhood region S xy .
α ≤ 1. The diffusion coefficient function g (∇ I ) is the same as that Eq. (4) can be rewritten by integrating the Eq. (14) in the diffu-
defined in Eq. (4). sion function as:
The modified diffusion model of CTAD-TFT has been reformu-
(i ) S xy  1
lated as g ∇ I t (x, y ), pt (x, y ) = (i ) S xy
, (15)
∇ I t (x, y )· pt (x, y ) 2
1+( )
1 
4
(i ) (i ) k
I t +1 (x, y ) = I t (x, y ) + g (∇ I t (x, y ) − ν (∇ I t (x, y ) where k is a positive fixed threshold used as a regulator to denoise
4
i =1 and also to improve the edge strength. Eq. (2) can be modified
× ∇ I t(i ) (x, y ) (11) as Eq. (16) using the proposed function of diffusion coefficient as
given in Eq. (15).
2.2.4. CTD model
The CTD model [12] was proposed to preserve edge regions and I t +1 (x, y )

1 
(i )
detail information during noise reduction process. The CTD model 4

S xy (i )
includes both local gradient and normalized variance of gray-level. = I t (x, y ) + g ∇ I t (x, y ), pt (x, y ) · ∇ I t (x, y ) (16)
The diffusion coefficient equation is given as 4
i =1

(i )  1 The new flux function of the proposed diffusion coefficient
g ∇ I t (x, y ), σt2, N (x, y ) = (12)
(i )
∇ I (x, y )·σt2, N (x, y ) 2 function given in Eq. (15) is formulated as follows:
1+[ t ]
k
S 
The normalized variance is defined as follows: φ ∇ I t(i ) (x, y ), pt xy (x, y )

(i ) S  (i )
σt2 (x, y ) − Min σt2 = g ∇ I t (x, y ), pt xy (x, y ) · ∇ I t (x, y ) (17)
σt2,N (x, y ) = 1 + · 254 (13)
Max σt2 − Min σt2 Fig. 3 illustrates the 3D plot of the diffusion coefficient function
(i ) S xy
2 2
where Min σ and Max σ are the minimum and maximum gray- of Eq. (15) with two variable functions ∇ I t (x, y ) and pt (x, y ).
t t
level variances of the diffused image at iteration t, respectively. It depicts that the values of the diffusion coefficient function
(i ) S xy
σt2 (x, y ) is the local gray-level variance (see Eq. (9)). g (∇ I t (x, y ), pt (x, y )) are increased suddenly when both the
116 M. Malarvel et al. / Digital Signal Processing 68 (2017) 112–126

location. The number of iterations t and k value of P–M, CTAD and

proposed model are t = 20, 50, 12 and k = 5, 0.05, 4, respectively,
where t and k are image dependent. The values of the parameters
t and k in the proposed model were taken dynamically using the
proposed adaptive selection procedure. The CTAD model incorpo-
rates the local variance in diffusion function, while the proposed
method incorporates local probability in diffusion function. Sig-
nificant changes in pixel values can be noticed in the proposed
methods in contrast to others.
To ensure the significant changes in pixel intensities, the P–M,
CTAD and proposed models are implemented on a chosen window
of noisy image indicated in Fig. 5(a). The average value (N 4 (x, y ))
of the diffusion coefficient of P–M, CTAD and proposed models on
the chosen window are 0.6712, 0.4954, and 0.9692, respectively. It
(i ) S xy
has been found that the larger diffusion coefficient value within an
Fig. 4. 3D plot of the flux function φ(∇ I t (x, y ), pt (x, y )) with variables
(i ) S xy
image region will bring the region smoother without loss of fine
∇ I t (x, y ) and pt (x, y ).
details [9] during diffusion process besides preserving the edge
details by increasing the pixel intensities in the proposed model.
(i ) S xy The pixel values of the sample window, which were considered
variables ∇ I t (x, y ) and pt (x, y ) are simultaneously decrease.
earlier from the input image, are processed using P–M, CTAD and
Hence the edges and fine details are preserved during smooth-
proposed models. The corresponding diffused values after the com-
ing process and thereby preciously preserve the defective regions.
(i ) S xy pletion of specific iterations in the respective diffusion processes
On the other hand, if both ∇ I t (x, y ) and pt (x, y ) are large, the
are shown side by side in the weld images of Fig. 5(b–d).
value of the diffusion coefficient function is come close to zero;
To explain the merit of the proposed diffusion model; P–M,
hence, the diffusion process stops and pixel values are preserved.
CTAD and proposed models are re-applied on the current diffused
Fig. 4 presents the 3D plot of the flux function of Eq. (17) with
(i ) S xy windows. The average values of the diffusion coefficients obtained
two variable functions ∇ I t (x, y ) and pt (x, y ). It illustrates that, from the diffused windows of the respective models are 0.9238,
(i ) S xy
when both ∇ I t (x, y ) and pt (x, y ) decrease concurrently, the 0.8988, and 0.9989. It has been found that the proposed model
(i ) S xy still produces higher diffusion coefficient value than others.
value of the flux function φ(∇ I t (x, y ), pt (x,
y )) will increase.
Hence, it provides sufficient smoothing for the noisy background Fig. 6 shows the relative comparison between possible prob-
S xy S xy
(i ) ability values of pt (x, y ) and the diffusion coefficient function
in the image. In contrast, when either ∇ I t (x, y ) or pt (x, y ) in-
creases and becomes quite large, the value of the flux function will
decrease to a number close to zero. It specifies that the proposed
diffusion model will stop the diffusion process for appropriate
edges and fine details in the image. Therefore, the modified dif-
fusion coefficient function and its flux function confirm that the
proposed diffusion model is removing noise appropriately, while
preserving both edges and fine details.
In order to prove the worthiness of the local probability in the
proposed diffusion function, the difference between operation per-
formances of P–M, local variance and the local probability value
in diffusion function on sample image is given in Fig. 5. Fig. 5(a)
is the noisy weld X-radiography image, which contains gas pore
defect with added Gaussian noise and its corresponding 3 × 3 win-
dow pixel values of sample location are chosen for comparison.
Fig. 5(b–d) illustrates the denoising results of P–M, CTAD and pro-
(i ) S xy S
posed model and their 3 × 3 window pixel values of the chosen Fig. 6. Graph of g (∇ I t (x, y ), pt (x, y )) with pt xy (x, y ).

Fig. 5. The comparison of various diffusion functions at sample 3 × 3 window: (a) added Gaussian noise image; (b) P–M model (iteration = 20 and k = 5); (c) CTAD model
with local variance (iteration = 50 and k = 0.05); (d) the proposed model with local probability (iteration = 12 and k = 5).
 M. Malarvel et al. / Digital Signal Processing 68 (2017) 112–126 117

S xy
p 1 (x, y ) is not considered as it is computed before the start of the
diffusion process. Thus, the lower probability value produces the
S xy
higher diffusion coefficient as shown in Fig. 6. Therefore, pt (x, y )
in the proposed diffusion coefficient can sufficiently smoothen the
noisy image.
In weld X-radiography images, the weld defects are of various
dimensions with low gradient in large area of noisy background
and thus need processing. The low gradient present in the fea-
ture space of a noisy image will affect the accuracy of detecting
the features viz. defects. Here, we ensure that the feature space
that is weld defect is retained without any disruption during the
S xy
smoothing process. pt (x, y ) increases the gradient value and de-
noises the feature space. Moreover, the random selection of k value
S xy
Fig. 7. Graph of pt (x, y ) in various iterations t for the image in Fig. 5(a). cannot always be effective for the noise removal, since the resul-
tant image may become over or under smoothened. Therefore, in
this proposed model, an automatic selection of k has been intro-
duced. The following section illustrates an adaptive selection of k
and besides, a criterion proposed to terminate the diffusion func-
tion iterations.

3.2. Adaptive selection of k and diffusion function iteration stopping

criterion

In most of the studies based on the classical AD model, the

diffusion function has been performed for a random number of
iterations to achieve the required denoising effect. In order to re-
duce the computational time, the number of iterations should be
minimized. To acquire the required denoised image, this paper in-
corporated the image entropy value and is considered in the stop-
Fig. 8. Shannon entropy value S t of image Fig. 5(a) for each iteration t.
ping criterion. The Shannon entropy method is employed in the
(i ) S xy stopping criterion, which is a well-known method that provides
g (∇ I t (x, y ), pt (x, y )) of the proposed model. It also shows
S xy S xy
information about disorder or variance in an image.
(i )
that g (∇ I t (x, y ), pt (x, y )) is decreasing while pt (x, y ) is in- Shannon entropy ( S t ) is calculated on the processed image in
creasing. For instance, the proposed model is applied on the im- each iteration (t ) as follows: in a given image, a spatial resolution
age given in Fig. 5(a) and the calculated value of M × N and bit level resolution b contains the pixels with gray val-
M −1  N −1 S xy S
[ pt (x, y )]/[( M − 2) · ( N − 2)] is denoted as pt xy (x, y ) ues ranging from 0 to L − 1, where L is the gray level of the pixel,
x=2 y =2
S xy namely, 2b . Let n (= M · N ) be the total number of pixels and ni be
at iteration t. pt (x, y ) of various iterations (t = 2 to 12) are de- the number of pixels with gray value i ∈ [0, L − 1]. Then the nor-
S xy (t )
picted in Fig. 7. It has been inferred that the value of pt (x, y ) malized histogram of the image at iteration t is H i = ni /n. The
is found to be decreasing with increasing of t. Here, the value of Shannon entropy at iteration t is defined as

Fig. 9. Smoothened images of Fig. 5(a) and 2D intensity profiles: (a) P–M model; (b) proposed model; (c) scan line 35 in Fig. 9(a); (d) scan line 35 in Fig. 9(b).
118 M. Malarvel et al. / Digital Signal Processing 68 (2017) 112–126
Fig. 10. Weld X-radiography images with different types of defects.
L −1
 (t )
(t )  Table 1
St = − H i ln H i (18)
i =0
3.2.1. Adaptive selection of k
Digitized images will have different entropy value based on its
existing noise levels. In this proposed model, the lower integral of
Shannon entropy,  S t =0  of the input image is initialized to k, a
threshold parameter in the diffusion coefficient function. The as-
signed k value is considered as constant threshold parameter value
and applied to entire iteration process during diffusion.
3.2.2. Diffusion function iteration stopping criterion
In order to terminate the iteration process to obtain the de-
sired smoothing during diffusion process, a stopping criterion is
proposed. The Shannon entropy value S t at iteration t is calculated
on diffused image. The diffusion function iteration is terminated if
| S t −1 − S t | < 0.0005.
Fig. 8 shows that the Shannon entropy value of each iteration
of Fig. 5(a) by the proposed model. It can be seen that the en-
tropy value of the processed image decreases in each iteration
and stabilizes after a finite number of iterations based on the
stopping criterion. In this experiment, diffusion process stops at
11th iteration where Shannon entropy values of the 10th iteration
and 11th iteration are equal to three decimal places of accuracy
(S 10 = S 11 = 4.008).
3.2.3. Performance analysis of adaptive selection of k and iteration
stopping criterion
In order to ascertain the appropriateness of adaptive selection
of k and the stopping criterion of the proposed model, a compari-
son analysis between the classical P–M and the proposed model is
experimented. The proposed model is applied on the image given
in Fig. 5(a) with adaptively selected threshold parameter k = 4 and
the process terminate at the 11th iteration and achieve the resul-
tant Signal-to-Noise Ratio (SNR) value 6.297. The same threshold
value has been used in the classical P–M model with 20 iterations
and obtained the SNR value 6.181. The corresponding smoothened
images are given in Fig. 9(b) and Fig. 9(a). The SNR is computed
using mean and standard deviation (S N R = μ/σ , where μ is the
mean and σ is the standard deviation of the considered pixel val-
ues). Higher the value of SNR, the better will be the noise level.
A scan line that passes through the defective region is consid-
ered from the smoothened images of both the models. Their 2D
intensity profiles of the scan lines are considered for visualization
and their SNR values are extracted to quantify and compare with
these models. These results are presented in Figs. 9(d) and 9(c).
The t and k parameter values.
Model No. of iteration (t ) Threshold parameter (k)
P–M 20 5
TCAD 150 5
CTAD 50 0.05
CTAD-TFT 30 Adaptive selection of k using
mean gradient magnitude
CTD 150 60
It has been found that a higher SNR value has been achieved in
the proposed model on the scan line with less iteration. Therefore,
the adaptive selection of k and diffusion stopping criterion of the
proposed model have performed well yielding smoothened image.
4. Image database, comparative experimentations and
discussions
4.1. Weld X-radiography image database
The X-radiography digital images of welds used in this paper
were taken from GRIMA X-ray database [27] collected by Domingo
Mery et al. in the interest of X-ray testing community. The test
images used for the analysis in this paper are taken from GRIMA
X-ray database and their acquisition standards are mentioned in
their website as follows: The X-radiography acquired according to
ISO 17636-1 and the optimized voltage is 150 kV for thickness of
8 mm to 20 mm steel. The weld X-radiography films have been
digitized with the Lumisys LS85 SDR scanner with 12 bit pixel
depth. The original 12 bit data was rescaled to 8 bits with a linear
LUT proportional to optical film density. The pixel size of the digi-
tized X-ray radiography image is 40.3 μm (630 dpi). The image file
format is TIFF without compression. The test images are 8 bit gray
values.
In order to focus our attention on the region where the weld
defects are available, we isolated a rectangular ROI manually that
contains the defect under study from the full size image. Fig. 10
shows some of the test images with defects such as porosity, gas
pore, undercut, slag inclusion, tungsten inclusion, and crack which
we have considered for experimentation. For performing the de-
noising experiment, Gaussian noises of different standard devia-
tions (chosen randomly between 10% and 40% of σ ) were added
to these images.
 M. Malarvel et al. / Digital Signal Processing 68 (2017) 112–126 119

Fig. 11. Smoothened results; Column (a), (b) and (c) first row: Noisy weld X-radiography images of Fig. 10(a), 10(b), and 10(c), respectively; from row II to VII: corresponding
smoothened images obtained from P–M, TCAD, CTAD, CTAD-TFT, CTD and the proposed model, respectively.

4.2. Implementation
The classical P–M [4], TCAD [7], CTAD [8], CTAD-TFT [9], and
CTD [12] models are implemented for smoothing the images, and
the outcomes of these models were compared with the proposed
model. Table 1 shows the various values of the parameters that
were taken in the existing models during implementation. Figs. 11
and 12 illustrates the various smoothened results obtained from
P–M, TCAD, CTAD, CTAD-TFT, CTD, and the proposed models. The
images presented in row I of Figs. 11(a), 11(b), and 11(c) & 12(a),
12(b) and 12(c) are the noisy images of Fig. 10(a)–(f), respectively.
The corresponding parametric values given in Table 1 for various
models are used to smooth the input images. The corresponding
resultant images are given from second to seventh row in Fig. 11
and Fig. 12.
The threshold parameter value (k) is adaptively taken in the
proposed model of the smoothened images in Figs. 11(a), 11(b),
11(c), 12(a), 12(b), and 12(c) at row VII are k = 4, 5, 5, 5, 4, and
5 respectively. Besides, the number of iterations t = 11, 19, 21, 20,
12, and 16 respectively. It has been revealed that the threshold
value of parameter k (obtained by the proposed adaptive selec-
tion method) taken to test all the images were less than or equal
to 5. Therefore, the proposed model selects the k value appro-
priately as ascertained by Section 2.1. Moreover the number of
iterations required to get the desired smoothing is less than or
close to minimum number of iterations performed in the existing
models besides the number of iterations required to stop the pro-
cess is also dynamically adaptive. We present the analysis of the
segmentation and quality performance of the smoothened images
in Sections 4.3 and 4.4.
4.3. Weld defect segmentation using six sigma concepts
On the smoothened image, a simple statistical method [9,28]
is applied to find the control limits. The upper and lower con-
trol limits for gray value variations in the smoothened images are
μ S I + C σ S I and μ S I − C σ S I respectively, where μ S I and σ S I rep-
resent the mean and standard deviation of the pixel values of the
smoothened image. C , a control constant, which is an integer in
the closed interval [1, 3]. During the image segmentation process,
if a pixel value lies within the control limits, then it is considered
as the background pixel. Otherwise, it is a foreground pixel. In this
120 M. Malarvel et al. / Digital Signal Processing 68 (2017) 112–126

Fig. 12. Smoothened results; Column (a), (b) and (c) first row: Noisy weld X-radiography images of Fig. 10(d), 10(e), and 10(f), respectively; from row II to VII: corresponding
smoothened images obtained from P–M, TCAD, CTAD, CTAD-TFT, CTD and the proposed model, respectively.

way an image gets segmented into two homogeneous regions. In ity. For the original image as F (i , j ) and the denoised image as
order to identify the defective region more preciously during the G (i , j ) with size M × N, MSE can be defined as
segmentation, we considered 2-sigma standard by choosing the

M −1 N
 −1
control constant C as 2. This segmentation procedure is followed 1  2
in all the experiments. Figs. 13–18 show the defect segmented re- MSE( F , G ) = F (i , j ) − G (i , j ) , (19)
MN
sults besides their 2D intensity profiles of a scan line that passes i =0 j =0

through the defective region. while the PSNR is defined as:

The segmented results and the 2D intensity profiles reveal that

the defects are segmented accurately since during smoothing pro-
cess, the defects also get preserved. Quantitatively, the proposed
method also yields higher SNR value where compared to other
methods.
4.4. Quality evaluation of smoothened images
The qualities of the smoothened images are quantified using
Mean Square Error (MSE), Peak Signal-to-Noise Ratio (PSNR), En-
tropy (E) and Mean Structural SIMilarity (MSSIM) [29].
The arithmetic dissimilarity between the original image and the
denoised image known as MSE, determines the loss of image qual-
2552
PSNR( F , G ) = 10 log10 . (20)
MSE( F , G )
A lower value of MSE indicates lesser error, and is directly
proportional to PSNR. However, the value of PSNR and error are
inversely proportional.
The entropy is a measure of the typical information of the im-
age distortion content. In Eq. (18), the Shannon entropy is utilized
to measure the entropy of the processed images, which illustrates
and quantifies the variance or disorder of the smoothened images.
Here, a lesser entropy value indicates lesser disorder.
MSSIM measures the image quality by computing the similarity
between images. SSIM is calculated between the two windows, F
 M. Malarvel et al. / Digital Signal Processing 68 (2017) 112–126 121

Fig. 13. Porosity weld defect segmentation and 2D intensity profiles of Fig. 11(a) using P–M, TCAD, CTAD, CTAD-TFT, CTD and the proposed models; (a)–(f) segmented results
and 2D intensity profiles of the line 75.

Fig. 14. Gas pore weld defect segmentation and 2D intensity profiles of Fig. 11(b) using P–M, TCAD, CTAD, CTAD-TFT, CTD and the proposed models; (a)–(f) segmented results
and 2D intensity profiles of the line 75.
122 M. Malarvel et al. / Digital Signal Processing 68 (2017) 112–126

Fig. 15. Undercut weld defect segmentation and 2D intensity profiles of Fig. 11(c) using P–M, TCAD, CTAD, CTAD-TFT, CTD and the proposed models; (a)–(f) segmented results
and 2D intensity profiles of the line 45.

Fig. 16. Slag inclusion weld defect segmentation and 2D intensity profiles of Fig. 12(a) using P–M, TCAD, CTAD, CTAD-TFT, CTD and the proposed models; (a)–(f) segmented
results and 2D intensity profiles of the line 45.
 M. Malarvel et al. / Digital Signal Processing 68 (2017) 112–126 123

Fig. 17. Tungsten inclusion weld defect segmentation and 2D intensity profiles of Fig. 12(b) using P–M, TCAD, CTAD, CTAD-TFT, CTD and the proposed models; (a)–(f)
segmented results and 2D intensity profiles of the line 55.

Fig. 18. Crack weld defect segmentation and 2D intensity profiles of Fig. 12(c) using P–M, TCAD, CTAD, CTAD-TFT, CTD and the proposed models; (a)–(f) segmented results
and 2D intensity profiles of the line 85.
124 M. Malarvel et al. / Digital Signal Processing 68 (2017) 112–126

Table 2
Performance measure of the smoothened weld images.

Weld image Metrics P–M TCAD CTAD CTAD-TFT CTD Proposed

Porosity MSE 15.8178 137.2364 57.3474 126.9496 108.5179 13.0955
PSNR 36.1393 26.7561 30.5457 27.0945 27.7758 36.9596
E 4.0925 4.4727 4.3221 4.0560 4.3014 4.0351
MSSIM 0.6240 0.0209 0.0954 0.5076 0.1514 0.6816

Gas Pore MSE 710.2416 982.9382 114.7859 1006.9900 952.2699 37.9825

PSNR 19.6167 18.2055 27.5319 18.1006 18.3432 32.3350
E 4.8834 4.9961 4.3497 4.2033 4.9413 4.0380
MSSIM 0.0085 0.0005 0.0650 0.3677 0.0007 0.7484

Undercut MSE 269.6776 542.7837 90.0423 555.7845 519.7502 29.3540

PSNR 23.8224 20.7845 28.5863 20.6817 20.9729 33.4541
E 4.8346 4.9825 4.7034 4.4808 4.8306 4.5038
MSSIM 0.0642 0.0017 0.0695 0.4073 0.0023 0.6976

Slag inclusion MSE 270.3212 533.1195 87.9503 558.3932 522.1584 31.5577

PSNR 23.8120 20.8626 28.6884 20.6614 20.9528 33.1393
E 4.9005 5.0266 4.7753 4.6392 4.9497 4.6168
MSSIM 0.0606 0.0017 0.0709 0.3931 0.0021 0.6603

Tungsten inclusion MSE 11.4969 111.7505 53.0739 102.9623 87.1567 10.8071

PSNR 37.5250 27.6483 30.8820 28.0040 28.7278 37.7937
E 3.5731 4.1048 3.9087 3.5253 3.8529 3.5406
MSSIM 0.7244 0.0285 0.1041 0.5635 0.1570 0.7505

Crack MSE 283.2189 553.3764 94.7641 598.0257 544.1375 68.3554

PSNR 23.6096 20.7006 28.3644 20.3636 20.7737 29.7831
E 4.4410 4.7699 4.0938 3.7312 4.7179 3.4077
MSSIM 0.0632 0.0016 0.0676 0.4162 0.0022 0.7189

and G, of common window size n × n. The common window size
is set as 3 × 3 in this study. A higher value of MSSIM indicates that
the resultant image has more similarity to the original image.
(2μ F μG + c 1 )(2σ F G + c 2 )
SSIM( F , G ) =
(μ2F + μ2G + c 1 )(σ F2 + σG2 + c 2 )
where μ F is the average of F ; μG is the average of G; σ F2 and σG2
are the variance of F and G respectively; σ F G is the covariance;
c 1 and c 2 are stabilizing parameters, generally c 1 = 0.05 and c 2 =
0.05. Recall that the covariance between F and G is defined as
L −1

1 Acknowledgments
σF G = ( F i − μ F )(G i − μG )
L−1
i =0
Their mean SSIM is defined as
1 
W
MSSIM( F , G ) = SSIM( F j , G j )
W image processing tool named “Bhadraloka” being developed for
j =1
where W is number of local windows of size 3 × 3 in the image.
Table 2 shows the performance measure of the proposed
method with existing ones on smoothened images. The perfor-
mance of the proposed method ensures the accuracy of image
smoothening.
5. Conclusions
In this paper, we have proposed a modified anisotropic diffu-
sion scheme based denoising, feature space gradient preserving
and weld defect detection for noisy weld X-radiography images.
The noisy defects in a weld X-radiography image are unclear and
have different sizes. In order to detect the weld defects, the noisy
image must be denoised or smoothened out and the fine details
must be preserved. The existing models have considered the pa-
rameters assigned randomly during diffusion process. Since the im-
age has noise, the defects remain unclear. Our diffusion model in-
corporates local probability value of gray levels into the traditional
diffusion coefficient function to denoising and defects preserving.
Furthermore, during the diffusion process, an adaptive threshold
parameter and automatic diffusion stopping criterion were intro-
duced based on the Shannon entropy value that also decides the
effectiveness of denoising, while preserving the fine details. The
(21) segmentation result of smoothened images show that the pro-
posed model can effectively remove the noises in noisy weld X-
radiography image. The denoising performance of the proposed
model was evaluated using MSE, SNR, PSNR, Entropy and MSSIM
metrics and the measurement results ensure that our model pro-
vides better results and outperforms those of others.
The authors would like to acknowledge the support rendered
by the managements of SASTRA University and Indira Gandhi Cen-
tre for Atomic Research, Kalpakkam by way of providing necessary
infrastructure for this research. This work was a part of generic
(22)
the research project entitled “Development of techniques for pro-
cessing radiographic images for automated detection of defects”
(No. 2013/36/40-BRNS/2305) in Dot NET platform at SASTRA with
the funding assistance from Board of Research in Nuclear Sciences
(BRNS), Department of Atomic Energy, Government of India.
Appendix A. Supplementary material
Supplementary material related to this article can be found on-
line at http://dx.doi.org/10.1016/j.dsp.2017.05.014.
References
[1] R. Ricardo, D. Mery, State-of-the-art of weld seam inspection by radiographic
testing, part I: image processing, Mater. Eval. 65 (2007) 643–647.
[2] J. Mohan, V. Krishnaveni, Y. Guo, A survey on the magnetic resonance image
denoising methods, Biomed. Signal Process. Control 9 (2014) 56–69.
[3] C. Pal, A. Chakrabarti, R. Ghosh, A brief survey of recent edge-preserving
smoothing algorithms on digital images, Proc. Comput. Sci. (2015) 1–40.
[4] P. Perona, J. Malik, Scale-space and edge detection using anisotropic diffusion,
IEEE Trans. Pattern Anal. Mach. Intell. 12 (1990) 629–639.
 M. Malarvel et al. / Digital Signal Processing 68 (2017) 112–126
[5] E. Catte, P.I. Lions, J.M. Morel, T. Coll, Image selective smoothing and edge de-
tection by nonlinear diffusion, SIAM J. Numer. Anal. 29 (1992) 182–193.
[6] L.I. Rudin, S. Osher, E. Fatemi, Nonlinear total variation based noise removal
algorithms, Physica D 60 (1992) 259–268.
[7] D.M. Tsai, S.M. Chao, An anisotropic diffusion-based defect detection for sput-
tered surfaces with inhomogeneous textures, Image Vis. Comput. 23 (2005)
325–338.
[8] S.M. Chao, D.M. Tsai, Astronomical image restoration using an improved
anisotropic diffusion, Pattern Recognit. Lett. 27 (2006) 335–344.
[9] S.M. Chao, D.M. Tsai, An anisotropic diffusion-based defect detection for low-
contrast glass substrates, Image Vis. Comput. 26 (2008) 187–200.
[10] D.M. Tsai, C.C. Chang, S.M. Chao, Micro-crack inspection in heterogeneously
textured solar wafers using anisotropic diffusion, Image Vis. Comput. 28 (2010)
491–501.
[11] S.M. Chao, D.M. Tsai, Anisotropic diffusion with generalized diffusion coefficient
function for defect detection in low-contrast surface images, Pattern Recognit.
43 (2010) 1917–1931.
[12] S.M. Chao, D.M. Tsai, An improved anisotropic diffusion model for detail- and
edge-preserving smoothing, Pattern Recognit. Lett. 31 (2010) 2012–2023.
[13] H. Tian, H. Cai, J.H. Lai, X. Xu, Effective image noise removal based on difference
eigenvalue, in: 18th IEEE Int. Conf. Image Process, 2011, pp. 3357–3360.
[14] V.B. Surya Prasath, D. Vorotnikov, Weighted and well-balanced anisotropic dif-
fusion scheme for image denoising and restoration, Nonlinear Anal., Real World
Appl. 17 (2014) 33–46.
[15] C.A.Z. Barcelos, M. Boaventura, E.C. Silva, A well-balanced flow equation for
noise removal and edge detection, IEEE Trans. Image Process. 12 (2003)
751–763.
[16] H. Luo, L. Zhu, H. Ding, Coupled anisotropic diffusion for image selective
smoothing, Signal Process. 86 (2006) 1728–1736.
[17] J. Yu, Y. Wang, Y. Shen, Noise reduction and edge detection via kernel
anisotropic diffusion, Pattern Recognit. Lett. 29 (2008) 1496–1503.
[18] D. Fang, Z. Nanning, X. Jianru, Image smoothing and sharpening based on non-
linear diffusion equation, Signal Process. 88 (2008) 2850–2855.
[19] Q. Chen, P. Montesinos, Q. Sen Sun, D. Shen Xia, Ramp preserving Perona–Malik
model, Signal Process. 90 (2010) 1963–1975.
[20] Z. Guo, J. Sun, D. Zhang, B. Wu, Adaptive Perona–Malik model based on the
variable exponent for image denoising, IEEE Trans. Image Process. 21 (2012)
958–967.
[21] Y.Q. Wang, J. Guo, W. Chen, W. Zhang, Image denoising using modified
Perona–Malik model based on directional Laplacian, Signal Process. 93 (2013)
2548–2558.
[22] B.J. Maiseli, H. Gao, Robust edge detector based on anisotropic diffusion-driven
process, Inf. Process. Lett. 116 (2016) 373–378.
[23] J. Xu, Y. Jia, Z. Shi, K. Pang, An improved anisotropic diffusion filter with semi-
adaptive threshold for edge preservation, Signal Process. 119 (2016) 80–91.
[24] I.B. Mhamed, S. Abid, F. Fnaiech, Weld defect detection using a modified
anisotropic model, EURASIP J. Adv. Signal Process. 46 (2012) 1–12.
[25] S.A. Halim, A. Ibrahim, Y.H. Manurung, PDE-based model for weld defect detec-
tion on digital radiographic image, Int. J. Signal Process. Syst. 1 (2013) 146–151.
[26] C. Dang, J. Gao, Z. Wang, F. Chen, Y. Xiao, Multi-step radiographic image en-
hancement conforming to weld defect segmentation, IET Image Process. 9
(2015) 943–950.
[27] D. Mery, V. Riffo, U. Zscherpel, G. Mondragón, I. Lillo, I. Zuccar, H. Lobel, M.
Carrasco, GDXray – The Grima Database of X-Ray Images, Department of Com-
puter Science, Universidad Católica de Chile, in Collaboration with Institute for
Materials Research and Testing (BAM), Berlin, 2015.
[28] S. Gopalakrishnan, S. Srinivasan, Border detection and cancer propagation
on spectral bands of malignant melanoma using six sigma threshold, in:
8th IEEE/ACIS Intl. Conf. on Computer and Information Science, ICIS, 2009,
pp. 586–592.
[29] O. Zahran, H. Kasban, M. El-Kordy, F.E.A. El-Samie, Automatic weld defect iden-
tification from radiographic images, NDT E Int. 57 (2013) 26–35.
Muthukumaran Malarvel received the M.C.A. de-
gree in 2007, the M.Phil. degree in Computer Science
in 2009, the M.Tech. degree in Computer Science and
Engineering in 2013 and currently pursuing Ph.D. de-
gree in Digital Image Processing at SASTRA Univer-
sity, India. From 2014 to 2017, he got fellowship as
a senior research fellow under Board of Research and
Nuclear Sciences, India project at SASTRA University,
India. He is currently working as Teaching Assistant
in Department of Computer Applications, School of Computing, SASTRA
University. His research interests include: image enhancement, image seg-
mentation, and statistical analysis & feature extractions in digital images.
He is also interested in developing scientific software applications.
125
Gopalakrishnan Sethumadhavan received the
M.Sc. and M.Phil. degrees in Mathematics from Bhara-
thidasan University, Tiruchirappalli, India, in 1985 and
1987 respectively and an M.Tech. degree in Quality
Assurance Technology from Regional Engineering Col-
lege, Tiruchirappalli in 1995. He has also acquired a
degree in Master of Computer Applications. He re-
ceived the Ph.D. degree from SASTRA University, India,
for his work in Biomedical Image Processing and Its
Applications to Clinical Diagnosis in 2006. He is a member of Computer
Society of India. He is currently a Professor of Department of Computer
Applications, School of Computing, SASTRA University. His research inter-
ests are in spatial domain image analysis with more emphasis on adaptive
segmentation and feature extraction. He also works in evolutionary algo-
rithms. He is especially interested in developing software in medical and
industrial applications using image processing and soft computing tech-
niques. He also bagged several best paper awards. He has over 30 years of
teaching experience and 15 years of research experience.
Purna Chandra Rao Bhagi joined Indira Gandhi
Center for Atomic Research, Kalpakkam after Metallur-
gical Engineering bachelor’s degree from Andhra Uni-
versity, India, and obtained Ph.D. from Saarland Uni-
versity, Germany. Currently, he heads Non-Destructive
Evaluation (NDE) Division working towards develop-
ment of advanced techniques for detection and quan-
titative evaluation of defects, residual stresses and mi-
crostructures by placing emphasis on sensors, numer-
ical modeling, image processing, quality assurance and automation. He is
responsible for development of comprehensive technologies for NDE of
steam generator tubes, clad tubes and main vessels of fast reactors. Dr.
B.P.C. Rao has a book and six book chapters and an Encyclopedia of Mate-
rials article to his credit. He is a recipient of Indian Nuclear Society Gold
Medal, National NDT Award and several best paper awards.
Soumitra Kar received Bachelor’s degree in Elec-
tronics and Telecommunication Engineering from Ja-
davpur University, Kolkata, with 1st class honors in
1985. He completed Ph.D. from CSE department of IIT
Bombay in 2002. He joined Bhabha Atomic Research
Centre (BARC), Mumbai and received Homi Bhabha
Award for securing 1st rank in Electronics discipline
in the orientation course in Nuclear Engineering at
BARC Training School (in 1986). Currently he is a sci-
entist in grade H at BARC, Mumbai, and holds the position of Head of
Image Processing and Machine Vision Section in EISD division. He is also
a Professor, Engg. Sciences, at Homi Bhabha National Institute (HBNI),
Mumbai. He has developed many techniques for face recognition, machine
vision for automation, and image analysis for inspection. His research area
of interest includes Image Processing, Artificial Vision and Biometrics. He
has over 30 years of research experience and over 14 years of teaching
experience.
Saravanan Thangavel received his Master’s degree
in Physics from Alagappa University, Karaikudi. He
completed his Ph.D. degree from Alagappa University
for his work on “Synthesis and Characterization of
substituted Lithium Manganese spinel oxide Materials
for Li-battery applications” in 2006. He joined Indira
Gandhi Centre for Atomic Research as scientific Of-
ficer in 2001. His research area of interest includes
Digital Real-time X-ray radiography and development
of image processing algorithms for nuclear applications. He has more
than 30 international journal publications. He has been awarded with
Group Achievement award from DAE for his contribution on PIE studies
on fast reactor fuel materials. He is a member of Indian Society of Non-
Destructive Testing and Indian Institute of Metals.
126 M. Malarvel et al. / Digital Signal Processing 68 (2017) 112–126

Arunmuthu Krishnan received the M.Sc. degree
in Physics from the Madras University, India, in 2000.
He has also acquired Master Diploma in Information
Technology in 2001. He joined Indira Gandhi Centre
for Atomic Research as scientific Officer in 2003. His
research interests include development of X-ray com-
puted tomography, image reconstruction using limited
projections, image and data processing, real-time pro-
gramming and computer vision. He was one of the people responsible for
the development and realization of the X-ray computed tomography sys-
tem at IGCAR. He selected as winner of The Ron Halmshaw Award 2010
for the best paper published in Insight on any aspect of industrial radio-
graphy or radiology. He is a member of Indian Society of Non-Destructive
Testing and Indian Institute of Metals.