International Journal of Signal Processing, Image Processing and Pattern Recognition

Vol.9, No.5 (2016), pp.209-218

http://dx.doi.org/10.14257/ijsip.2016.9.5.18

A Review on Image Segmentation with its Clustering Techniques

Priyansh Sharma1 and Jenkin Suji2

Dept. of EC, ITM University
[email protected], [email protected]

Abstract
Segmentation refers to a technique in which an image in digital form is partitioned into
multiple segments (basically groups of pixels, also termed as Super pixels). This paper is
a survey on Image Segmentation with its clustering techniques. Image Segmentation is the
procedure of apportioning a picture into numerous segments, to change the
exemplification of a picture into another which is more useful and easy to segment. A few
universally useful calculations and approaches have been developed for picture division.
It separates a digital picture into numerous locales to investigate them. It is likewise used
to recognize segment items in the picture. A few picture segmentation procedures have
been developed by the specialists with a specific end goal to make pictures smooth and
simple to access. This paper describes segmentation techniques, advantages and
disadvantages of the clustering methods and a comparison of the techniques.

Keywords: Image segmentation; K-Means algorithm; Fuzzy C-Means algorithm; log –

based clustering

1. Introduction
A digital image is basically an array in two dimensions of square regions known
as pixels. In case of a monochrome or gray-scale image, the intensity of every pixel
is denoted by a numeric value. It is also depicted as a fixed set of values in digital
form of a two-dimensional image called pixel elements or pixels. Digital images can
be formed by a number of input devices and techniques: Scanners, Digital cameras,
Coordinate measuring machines. Digital images are of various types. They are
Binary image, Grayscale image and Color image.
A Binary image is basically a digitalized image which has only two values, 0 and
1, for each pixel. They are also termed as two-level or bi-level. A binary image is
generally stored in memory space as a bitmap, a packet of bits array. Binary images
often ascend in digital image processing as masks or as the product of certain
operations such as segmentation and image thresholding.
Gray-scale images typically have values in the array from 0 and going to 255,
where 0 shows the black color, 255 shows white color and the remaining values in
between 0 and 255 represents shades of gray.
A color image is denoted in an array in two dimensions of Green, Red and Blue
triples. Basically in a triple, each number lies in between 0 to 255, where 0 shows
that none of the prime color is there in that pixel and 255 shows a maximum amount
of that main color in that pixel.
Image processing is the technique in which a picture is converted into digital
form and to perform some processes on it, in order to acquire a superior image or to
extract some essential data from it. It is basically a kind of signal exemption in
which input is image, such that a photograph or video frame and output can be
image or features related with that image. Usually Image Processing system
comprises of treating images as signals in two dimensions while applying already
usual signal processing methods to them.

ISSN: 2005-4254 IJSIP

Copyright ⓒ 2016 SERSC
International Journal of Signal Processing, Image Processing and Pattern Recognition
Vol. 9, No. 5 (2016)

2. Image Segmentation
Image Segmentation is defined as the method in which an image is partitioned
into many parts, such that an image is depicted into something that is easy to
express and easy to study. Image segmentation is vital for meaningful analysis and
interpretation of medical images [1]. Segmentation is the advanced technique in
which a digitalized image is partitioned or segmented into numerous segments or
parts based on the values of pixel. It is a perilous and important section of image
exploration system [2].
The purpose of Image Segmentation is to divide an image into semantically
interpretable regions with regard to a particular application and to identify
homogeneous regions within the image as discrete and belonging to distinct objects
[1].
There are several algorithms and methods that have been established for
segmenting image. Modern medical imaging modalities like MRI and CT scans
generate large images which cannot be studied manually. This develops the
requirement for more effective and robust image determination approaches, tailored
to the problems met in medical images [3].

3. Segmentation Technique
There are many methods for segmenting an image that have been recognized by
scientists and researchers. Hence, there are several such techniques that are quite
popular, important and are regularly used for image segmentation. These are
classified as follows:
 Region Based Segmentation
 Edge Based Segmentation
 Threshold Segmentation
 Clustering Based Segmentation
Region Based Segmentation systems try to group pixels together with identical
features (such as estimated gray level quality) into regions [5]. This segmentation
technique is quite easy and efficient as compared to other techniques and it is also
noise resilient. It is a type of segmentation technique in which an image is split into
many unlike sections based on criterion i.e. object, color or intensity. The
segmentation methods based on region are characterized into the following
categories i.e. region splitting, region growing, and region merging [4].

Segmentation Techniques

Pixel-Based Edge-Based Region-Based

Edge Region
detection growing
thresholding Clustering

Gradient Split /
Otsu K-means mode Merge

Global Active Graph Cut

Fuzzy contours (hybrid)
C-mean
Adaptive
Level
sets

Figure 1. Image Segmentation Technique

210 Copyright ⓒ 2016 SERSC

 International Journal of Signal Processing, Image Processing and Pattern Recognition
Vol. 9, No. 5 (2016)

Edge based segmentation technique is one of the most basic step used in image
segmenting process. In edge detection technique, the image is split by spotting the
difference in pixels of the digital image or intensity [4]. Edge detection technique is
determining the value of pixels on the boundaries of region. The image
segmentation is done by edge detection method by noticing pixels or edges in
between diverse sections that basically have quick change in intensity are removed
[2] and coupled to create closed margins of object. The outcome is an image in
binary form [6]. Gray histogram and Gradient are two primary techniques which are
used for segmenting image via edge detection [4].
Thresholding technique is the primary method used for image segmentation. It is
a method used for discriminating background from foreground. The gray scale
image can be converted into binary image by choosing an acceptable threshold value
T. The image which is in binary form must have all the vital knowledge of the shape
and position of the interested objects (foreground)[7]. The algorithms based on
thresholding can be obtained physically by having some priori knowledge or
repeatedly by formation of image [8].

4. Clustering Algorithm
Clustering in image processing is basically defined as the technique in which
groups of identical image primitive are identified [1].Clustering is a method in
which objects are unified into groups based on their characteristics. A cluster is
basically an assembly of objects which are similar between them and are not similar
to the objects fitting to additional clusters. An image can be segmented based on its
keyword (metadata) or its content (description)[9].
Data clustering algorithms are built over the whole image and these algorithms
studies data distance. The pixels of a cluster are not surely connective in data
clustering. The algorithms based on clustering are further categorized into two
techniques i.e. Hierarchical and Partitional. Hierarchical clustering is a successive
split process, which fallouts in a cluster structure which is hierarically nested, and
partitional clustering is an iterative partitioning process [10]. The clustering
algorithms are as follows :

4.1. K-Means Clustering

K-Means algorithm is a quite known clustering algorithm that categorizes the
input points of data into many groups produced on their inherent distance from one
another [10].
The K-means algorithm [2], is one of the most common iterative algorithm which
is basically used for its simplicity of implementation and convergence speed. K -
means also produces moderately high quality clusters seeing the low level of
computation and calculations required. The K-means technique focuses to diminish
the total of squared distances in between all points of pixel and the center of the
cluster. K-means algorithm is statistical clustering algorithm.
In this algorithm, the number of partitions are basically predefined. The cluster
centers are arbitrarily initialized for already defined clusters. Each data point is then
allocated to one of the nearest cluster. The cluster centers are then basically re -
estimated which and new centroid is computed. Until there is no significant change
in the cluster centre, this method is repeated [11].
K-means algorithm is particularly based over the index of resemblance or
difference between sets of components of data. K-means algorithm is
nondeterministic, iterative, numerical and a method which is not supervised [4]. K-
means performs good with many data sets, but its decent presentation is limited
mainly to compact groups.

Copyright ⓒ 2016 SERSC 211

International Journal of Signal Processing, Image Processing and Pattern Recognition
Vol. 9, No. 5 (2016)

The popular K-means algorithm [11] is an error minimization algorithm where

the function to minimize is the sum of squared error:
( ) ∑ ∑ ( )2 (1)
Where is the centroid of cluster and K the number of clusters (known a priori).
Two factors have made the k-Means popular: it has linear time complexity and its
easy implementation.
Advantages:
1) This algorithm is quite easy to understand, robust and fast.
2) It is relatively well organized or systematic algorithm: O (tknd), where n is
the number of objects, k is the number of clusters, d is the dimension of each
object, and t is the iterations. Basically k, t, d << n.
3) When the data set is discrete, this algorithm offers the best result.

4.2. Fuzzy C-Means Clustering

The Fuzzy C-Means algorithm (normally abbreviated as FCM) is basically an
iterative algorithm[1] that helps to find clusters in data and in which the idea of
fuzzy membership is used. As an alternative of giving a pixel to a single cluster,
each pixel can have distinctive membership values on each cluster[12].Fuzzy C-
means(FCM) is a system of clustering which lets one piece of data to belong to two
or extra clusters. This approach (developed by Dunn in 1973 and elevated via
Bezdek in 1981) is most often used in recognition of pattern.
This algorithm is an unverified clustering algorithm that is functional to quite
many issues which involves classifier and clustering design, feature analysis. This
algorithm has a number of applications like astronomy, geology, image analysis,
chemistry, shape analysis, medical diagnosis, and recognizing of the target. The
Fuzzy C-Means [2] tries to catch clusters of pixel in the data by reducing the
objective function as shown in the equation below:
∑ ∑ | |2 (2)
Here K is the objective function. After one iteration of the algorithm, the value of
K becomes small. It shows that the algorithm is coming closer to a fair separation of
pixels into clusters. In the image, N represents the number of pixels in the algorithm,
C represents the clusters number, here fuzziness factor is denoted by m (a value
always greater than 1), the p th pixel in N is denoted by x p , the q th cluster in C is
denoted by c q and | | is the Euclidean distance between x p and c q[12].
Many feature spaces can represent an image, and the FCM algorithm categorizes
the image by making groups of identical data points in the feature space into
clusters. This algorithm is obtained by iteratively reducing a cost function which is
reliant on the displacement of the pixels to the centre of the clusters. The pixels on
an image in the immediate neighborhood own the same feature data i.e. the pixels on
image are correlated. Therefore, the association of neighboring pixels is an essential
property that is of huge importance in image segmentation.
Advantages:
1) For the partly cover set of data, this algorithm gives good results and it is
quite efficient than k-means algorithm.
2) The data point completely belong to single centre of cluster, where as in
FCM the data point is given membership to each centre of cluster, because
of which data point belongs to more than one centre of cluster.

212 Copyright ⓒ 2016 SERSC

 International Journal of Signal Processing, Image Processing and Pattern Recognition
Vol. 9, No. 5 (2016)

4.3. Log-Based Clustering

The logs which are maintained by information retrieval process (like web servers
access logs) can be used to achieve image segmentation using clustering. But log–
based clustering can result in incomplete information. This is because log –based
clusters are created on those forms which are accessed by some of the users. To
solve this issue, a log –based vector is kept for each session vector which is built on
log–based documents [13]. A given document in a hybrid matrix form is usually
represented by one vector. And for those documents that are in a log based
document is created in a hybrid matrix form. Also the unaccessed document
generates their own vector. The documents which are a cluster of content based
clustering algorithm is basically represented by a hybrid matrix. The log–based
clustering algorithm is fairly precise in comparison to previous algorithm but this
algorithm is not very proficient to use.
An image can be clustered by the recovery systems which are managed by the
information reclamation procedure. The session keys are retrieved and shaped for
retrieval process. By this process the session clusters are made. Each session cluster
creates documents based on log and resemblance of image couple is recovered. For
each session vector a log-based vector is formed which depends on the documents
based on log. Further, this vector is swapped by the session vector. The unaccessed
document generates its own vector. A hybrid matrix is created with a single distinct
document vector and one log-based clustered vector. Finally the hybrid matrix is
clustered. This method is quite tough to achieve in the case of multidimensional
images. To solve this problem, hierarchical clustering is approved.

5. Literature Survey
SyojiKobashi (2013) et al. present that in order to properly study the medical
images, the study of image segmentation is essentially required. Though there are
numerous literatures that studied automated image segmentation, it is then also quite
tough to section regions of interest in any form of picture. In this paper two
methodologies of interactive image segmentation method based on Fuzzy
Connectedness Image Segmentation (FCIS) are introduced in order to shorten the
process time and to reduce the effort of users. The first technique enhances object
affinity of FCIS in line with user’s additional seed voxels. The second method
samples the outline of the object affinity by making use of the radial-basis function
network (RBFN). These suggested techniques enhances segmentation outcomes for
the other miss classified voxels and for the primary seed voxels. The second
technique however produced better outcomes than the first according to
experimental outcomes [14].
Jay C. Acharya (2013) et al. present that the phenomena like atmosphere, water,
temperature and pressure have a direct impact on the properties of submerged
images. This tells the significance of Image segmentation, in which a digitalized
image is segmented into numerous segments or regions based on the value of pixels.
Image Segmentation is the most advanced technique among all essentially
computerized image recognition systems. There exists many such segmentation
techniques that are used for segmenting submerged images. The presentation of an
algorithm performing Image Segmentation is based on the interpretation of image.
Algorithms based on Image segmentation namely adaptive image thresholding,
Fuzzy C means with thresholding, K-means segmentation, edge based image
segmentation are applied for submerged images and these techniques are also
compared on the parameters like Relative Entropy, Redundancy and Discrete
Entropy. Fuzzy c means with thresholding (FCMT) performs far better than other
methods is shown by the experimental results is shown in the paper[15].

Copyright ⓒ 2016 SERSC 213

International Journal of Signal Processing, Image Processing and Pattern Recognition
Vol. 9, No. 5 (2016)

AC Bharath (2014) et al. present that with a huge number of images, User
assisted segmentation of lungs parenchyma pathology bearing regions becomes
more tough. In this paper, a technique using Bayesian network model based
(BNMB) Image Segmentation is proposed which is a graphical model for
segmenting lung tissues from the CT images of chest. The aim of this technique is to
partition of lung parenchyma from the rest of chest CT image. This is done with the
use of a probabilistic graph construction to show the coordination between edge
segments and super pixel regions. By using an iterative method which depends on
the probability model, the regions are identified and then finally regions are merged
together. In this paper, experimental results demonstrates high precision and
productivity in segmenting CT image of lungs [16].
Xuemei Cui (2014) et al. present that on weak edges the watershed transform
shows good performance but in process of image segmentation not able to get useful
partition in an image. In this work an enhanced watershed algorithm present for
image segmentation. By the use of watershed algorithms get the marks on the
forefront and background objects obtain better partition. With properties of gradient
image and morphological open close function reconstruction operations the
watershed transform algorithm applying to obtain the structure tag. Then, obtained
tag using morphological minimum calibration technology forces as an original local
minimum value of the gradient image, and shields the original all local minimum
value in the gradient image. At the end, watershed which is enhanced by the
gradient images will have a image segmentation, in such aspects as outline eliminate
over segmentation and regional positioning have a very good segmentation effect
[17].
Yongjing Wan (2015) et al. present that Image edge detection is sensitive to noise
which is contained by natural images so that it affects the quality of the picture
partition. In order to remove noise and improve edge detection accuracy, then
improving the quality of picture segmentation, a novel image se gmentation
algorithm via neighborhood the main section analysis and Laplace operator is
proposed. The feature vectors of each pixel are extracted through the main section
analysis to get the main component, which effectively suppresses noise. Then the
Laplace operator is used to detect the edge to realize the image segmentation.
Compared to the traditional image segmentation of Sobel operator and LOG
operator, the algorithm is proposed estimate the parameter values by principal
component analysis instead of depending on experience. It can effectively decrease
the noise on the image interference and simplify the computational complexity.
Experimental results demonstrates that the algorithm can effectively improve the
segmentation of the picture with a strong advantage in the accuracy and robustness
[19].
Junpeng Wu (2015) et al. present that Markov Random Field (MRF) model along
with Image Segmentation is anticipated to attain the sonar image segmentation. In
this firstly, the assets of the field and the label field of sonar image is recognized by
the MRF model. After this, the remaining image segmentation is handled by region
growing centered on the primary image segmentation and finally this minimizes the
segmentation signs which are a result of MRF method. In this paper, the
experimental results demonstrates that the discussed technique obtains image
segmentation well.
Nupur J. Gandhi (2014) et al. present that Image segmentation is an important
and interesting matter in an image study. In an image, segmenting object is quite
tough and costly. A region based Image Segmentation technique that is mean shift
clustering approach is studied in this paper. In Image Segmentation, the most
essential phase is pre-processed image by a typical mean shift based segmentation
that conserves needed gaps which are there in the image and guarantees over

214 Copyright ⓒ 2016 SERSC

 International Journal of Signal Processing, Image Processing and Pattern Recognition
Vol. 9, No. 5 (2016)

segmentation in the picture. This mean shift segmentation process which partitions
the regions decreases the noise sensitivity and therefore increase the presentation of
segmentation [21].

Table 1. Monochrome Image Segmentation Techniques [22]

Segmentation Method Advantages Disadvantages

Technique Description

Histogram This It normally 1) Basically for an

Thresholding segmentation doesn’t require image without some
process a previous noticeable peaks or
mainly needs information of the image with wide
that the the image. And and even valleys,
histogram of it includes this technique does
the picture fewer not works fine.
has numerous calculational
2) This technique
peaks, where complexity.
does not reflect the
each relates to
facts, so it is not
a specific
able to assure that
region.
partitioned regions
are contiguous.
Region-Based Group Pixels It works best 1) The biggest
Approaches into when the drawback is that
homogeneous region this technique is
regions. homogeneity quite costly both in
Includes standard is memory
region quite simple to management and
merging, describe. This time used in
region technique first computations;
splitting, removes noise
2)This technique
region and then
has a characteristic
growing or detects the
dependency on the
their edges.
variety of seed
combinations
region and the way
in which regions
and pixels are
studied;

Copyright ⓒ 2016 SERSC 215

International Journal of Signal Processing, Image Processing and Pattern Recognition
Vol. 9, No. 5 (2016)

Based on the Edge detection 1)It does not work

recognition of technique is the fine with images in
Edge Detection
discontinuity, method in which the edges are
Approach
basically which human not properly-
efforts to notices things defined or there are
locate points and works well many edges;
with more or for images
2) It is not a
less abrupt having fine
insignificant job to
changes in difference
crop a closed curve
gray level. between
or boundary;
regions.
3) Less resistant to
noise than other
techniques.

Fuzzy Apply fuzzy Fuzzy 1)The

Approaches operators, membership determination of
mathematics, function can be fuzzy membership
used to is not a small job;
Properties and
represent the
inference 2) The computation
degree of some
rules, provide involved in fuzzy
properties or
a way to approaches could
linguistic
handle the be intensive.
phrase, and
uncertainty
fuzzy IF-
inherent in a
THAN rules
diversity of
can be used to
problems due
perform
to ambiguity
approximate
rather than
inference.
randomness.

Neural Network Using neural No need to 1) Training time is

Approaches networks to write long;
perform complicated 2)Initialization may
classification programs. Can effect the result;
or clustering fully utilize the 3)Overtraining
parallel nature should be avoided.
of neural
networks.

216 Copyright ⓒ 2016 SERSC

 International Journal of Signal Processing, Image Processing and Pattern Recognition
Vol. 9, No. 5 (2016)

Table 2. Image Segmentation Methods: A Comparative Study [23]

Parameter Threshold Based Edge Based Region Based

Segmentation Segmentation Segmentation

Nature of the Black-White Black-White Black-White

Output Image
Spatial Neglected Neglected Considered
Information

Region- Moderate Moderate High

Continuity

Computation Less Moderate High

Complexity

Speed Fast Moderate Slow

Noise Noise Noise Less
Immunity Immunity Immunity
Detection of Poor Poor Moderate
Multiple
objects
Automaticity Interactive Interactive Interactive
(Semi Automatic) (Semi
Automatic)
Accuracy Moderate Moderate Good

6. Conclusion
In mainframe vision, image segmentation is defined as the process in which an
image is partitioned into many sections. The aim of segmentation is to solve and/or
vary the depiction of an image into something that is more essential and easy to
examine. Image segmentation is basically used to see objects and background in
images. More exactly, the process in which a label is assigned to every pixel in an
image so that pixels with exactly same label share certain visual properties is called
Image Segmentation. Image segmentation is an essential signal handling tool that is
widely used in many applications including object base coding, object detection,
object tracking, image retrieval, and clinical organ or tissue identification. Image
segmentation is very important in many medical imaging applications. Here
presenting some survey and review of the current technologies and approaches for
the segmentation of images. In this paper some important issues in image
segmentation have been discussed. Reviews related with image segmentation and
problem identification about those papers have been covered.

References
[1] H. M. Moftah, A. TaherAzar, E.T. Al-Shammari, N. I. Ghali, A. E. Hassanien and M. Shoman,‖
Adaptive k-means clustering algorithm for MR breast image segmentation‖, Neural Comput & Applic
vol. 24, (2014), pp.1917–1928.

Copyright ⓒ 2016 SERSC 217

International Journal of Signal Processing, Image Processing and Pattern Recognition
Vol. 9, No. 5 (2016)

[2] M. Xess and S. A. Agnes,‖ Survey On Clustering Based Color Image Segmentation And Novel
Approaches To Fcm Algorithm‖, IJRET: International Journal of Research in Engineering and
Technology, pp. 346-349.
[3] K. Shrivastava, N. Gupta and N. Sharma,‖ Survey Paper On Image Segmentation Using K-Means
Clustering‖, International Journal of Advanced Technology & Engineering Research (IJATER), vol. 4,
no. 5, (2014) Sept., pp: 8-11.
[4] W. Khan, ―Image Segmentation Techniques: A Survey‖, Journal of Image and Graphics, vol. 1, no. 4,
(2013) December, pp. 166-170.
[5] M. Lalitha, M. Kiruthiga and C. Loganathan, ―A Survey on Image Segmentation through Clustering
Algorithm‖, International Journal of Science and Research (IJSR), vol. 2, no. 2, (2013) February, pp.
348-358.
[6] M. S. Sonawane and C.A. Dhawale,‖ A Brief Survey on Image Segmentation Methods‖, International
Journal of Computer Applications (0975 – 8887) National conference on Digital Image and Signal
Processing, DISP, (2015).
[7] M. J. Kumar, G. Raj Kumar and R. Vijay Kumar Reddy, ―Review On Image Segmentation Techniques‖,
International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 – 0882
vol. 3, no. 6, (2014) September, pp. 992-997.
[8] S. Agrawal and D. K. Xaxa,‖ Survey on Image Segmentation Techniques and Color Models‖, (IJCSIT)
International Journal of Computer Science and Information Technologies, vol. 5, no. 3, (2014), pp.
3025-3030.
[9] S. Bhowmik and V. Datta, ―A Survey on Clustering Based Image Segmentation‖, International Journal
of Advanced Research in Computer Engineering & Technology, vol. 1, no. 5, (2012) July, pp: 280-284.
[10] D. Naik and P. Shah,‖ A Review on Image Segmentation Clustering Algorithms‖, (IJCSIT)
International Journal of Computer Science and Information Technologies, vol. 5, no. 3, (2014), pp. 3289
– 3293.
[11] A. Gulhane, P. L. Paikrao and D. S. Chaudhari, ―A Review of Image Data Clustering Techniques‖,
International Journal of Soft Computing and Engineering (IJSCE) ISSN: 2231-2307, vol. 2, no. 1,
(2012) March, pp: 212-215.
[12] P. TamijeSelvy, V. Palanisamy and T. Purusothaman, ―Performance Analysis of Clustering Algorithms
in Brain Tumor Detection of MR Images‖, European Journal of Scientific Research ISSN 1450-216X
vol. 62, no. 3, (2011), pp. 321-330.
[13] Agarwal, ―International Journal of Advanced Research in Computer Science and Software Engineering‖,
vol. 3, no. 4, (2013) April, pp. 764-766.
[14] S. Kobashi, K. Kuramoto and Y. Hata, ―Interactive fuzzy connectedness image segmentation for
neonatal brain MR image segmentation‖, 2013 IEEE International Conference on Systems, Man, and
Cybernetics, pp. 1799-1805.
[15] Jay C. Acharya, Sohil A. Gadhiya and Kapil S. Raviya,‖ Objective Assesment Of Different
Segmentation Algorithm For Underwater Images‖, 4th ICCCNT –(IEEE), (2013).
[16] A. C. Bharath and D. Kumar, ―An Improved Bayesian Network Model Based Image Segmentation in
Detection of Lung Cancer‖,International Conference on Recent Trends in Information Technology
IEEE. (2014).
[17] X. Cui, G. Yang, Y. Deng and S.Wu,‖ An Improved Image Segmentation Algorithm Based on The
Watershed Transform‖, IEEE, (2014), pp. 428—431.
[18] O. C. Linares, G.Botelho, F. Rodriguesand and J. B. Neto, ―An Adjustable Error Measure for Image
Segmentation Evaluation‖, 28th SIBGRAPI Conference on Graphics, Patterns and Images, pp. 80- 86,
(2015).
[19] Y.Wan, T. Lu, W.Yang and W.Huang, ―A Novel Image Segmentation Algorithm via Neighborhood
Principal Component Analysis and Laplace Operator‖, 2015 International Conference on Network and
Information Systems for Computers IEEE, pp. 273- 276.
[20] J. Wu and H. Guo, ―A Method for Sonar Image Segmentation Based on Combination of MRF and
Region Growing‖, Fifth International Conference on Communication Systems and Network
Technologies IEEE, (2015), pp. 457- 460.
[21] N. J. Gandhi, V. J. Shah, Jr. and R. Kshirsagar, ―Mean Shift Technique for Image Segmentation and
Modified Canny Edge Detection Algorithm for Circle Detection‖, International Conference on
Communication and Signal Processing, India, (2014) April 3-5, pp. 246- 250.
[22] S. Banu, ―The Comparative Study on Color Image Segmentation Algorithms,‖ International Journal of
Engineering Research and Applications (IJERA), vol. 2, no. 4, (2012) July-August, pp.1277-1281.
[23] A. M. Khan and S. Ravi, ―Image Segmentation Methods: A Comparative Study‖, International Journal
of Soft Computing and Engineering (IJSCE), ISSN: 2231-2307, vol. 3, no. 4, (2013) September.

218 Copyright ⓒ 2016 SERSC