A Review on Image Segmentation using Clustering and Swarm Optimization Techniques

IJSRD - International Journal for Scientific Research & Development| Vol. 3, Issue 10, 2015 | ISSN (online): 2321-0613
All rights reserved by www.ijsrd.com 803
A Review on Image Segmentation using Clustering and Swarm
Optimization Techniques
Pragya Sharma1
Unmukh Datta2
1,2
Department of Computer Science & Engineering
1,2
MPCT Gwalior, India
Abstract— The process of dividing an image into multiple
regions (set of pixels) is known as Image segmentation. It
will make an image easy and smooth to evaluate. Image
segmentation objective is to generate image more simple
and meaningful. In this paper present a survey on image
segmentation general segmentation techniques, clustering
algorithms and optimization methods. Also a study of
different research also been presented. The latest research in
each of image segmentation methods is presented in this
study. This paper presents the recent research in biologically
inspired swarm optimization techniques, including ant
colony optimization algorithm, particle swarm optimization
algorithm, artificial bee colony algorithm and their
hybridizations, which are applied in several fields.
Key words: ACO, PSO, Image segmentation, K-Mean
Fuzzy C-Means
I. INTRODUCTION
The principle of image segmentation is to division an image
into particular regions with respect to a suitable application.
Essential step is used for image analysis in Image
segmentations are object visualization, object representation
and several other image processing jobs. In image
segmentation, segmentation is stand on measurements in use
from the image and might be grey level, color, texture,
motion or depth. At first, Segmentation partitions an image
into its essential objects or regions. Segmentation permits in
extracting the objects in images. In image processing
segmentation of images is not an easy job. Segmentation is
an unsupervised learning. Model based object extraction is a
supervised learning, e.g., template matching. After a
successful segmenting the image, the contours of objects can
be extracted using edge detection and/or border following
techniques. Shapes of objects are based on texture, shape
and color objects can be recognized. Image segmentation
methods are broadly used in comparison searches.
Segmentation algorithms are supports on one of
two essential properties of color, texture or gray values:
similarity and discontinuity. First category is based on
partitioning an image into regions that are similar, according
to predefined criteria. Histogram thresholding approach falls
under this category. Second category is to partition an image
based on abrupt changes in intensity, such as edges in an
image.
Generally image segmentation is a primary and vital
step in a sequence of processes aimed at overall image
understanding. Applications of image segmentation contain:
 Identifying objects in a picture for object-based
measurement like as shape and size.
 Identifying objects which are at dissimilar distances
from a sensor using depth dimensions from a laser range
finder enabling path planning for mobile robots.
 In object-based video compression (MPEG4) objects
identifying in a moving scene.
Since a technique applied to single image may not
remain doing well to other type of images, therefore
segmentation methods have been separated into three
categories, i.e. segmentation techniques support on classical
method, hybrid techniques and AI techniques. A number of
the most well-known image segmentation methodologies
including Edge based segmentation, Fuzzy theory based
segmentation, Partial Differential Equation (PDE) based
segmentation, Artificial Neural Network (ANN) bases
segmentation, threshold based image segmentation, Region
based image segmentation and threshold based image
segmentation are highlighted in figure. Fig.1 enclosed main
and well-known image segmentation techniques used for the
purpose of image segmentation.
II. IMAGE SEGMENTATION TECHNIQUES
Various image segmentation techniques have been expanded
by researchers and scientists, some of the most important
and commonly used image segmentation techniques are
shown in Fig.1. Latest research work on image segmentation
techniques highlighted in Fig.1 is discussed and evaluated
below.
Fig. 1: Techniques of Image Segmentation
A. Threshold Based Image Segmentation
Thresholding is an old, simple and popular technique for
image segmentation. By thresholding Image segmentation is
easy but controlling approach for segmenting images having
light objects on dark background. By the Thresholding
operation a multilevel image convert into a binary image
i.e., it select a proper threshold T, to split image pixels into
separate objects and several regions from background.
Some pixel (x, y) is measured as a part of object if
its intensity is greater than or equal to threshold value i.e.,
f(x, y) ≥T, else pixel belong to background. There are two
types of thresholding methods. They are classified as local
and global thresholding. If T is constant then it is known as
global thresholding otherwise it is local thresholding.
Global thresholding techniques can fail when the
background illumination is uneven. In neighboring
thresholding, multiple thresholds are used to compensate for
uneven illumination. There are some disadvantages of
thresholding method. It generates only two classes, and it
cannot be used to multichannel images. Thresholding does
not take into account the spatial characteristics of an image
so it is sensitive to noise. This corrupts the histogram of the
image, making separation more difficult[1].

A Review on Image Segmentation using Clustering and Swarm Optimization Techniques
(IJSRD/Vol. 3/Issue 10/2015/175)
B. Region Based Image Segmentation
This segmentation is easy as judge against to other methods
and also noise resilient. It partitions a picture into distinctive
areas taking into account pre-characterized criteria, i.e.,
color, intensity, or object. Region based segmentation
technique are ordered into three principle classifications,
i.e., region growing, region splitting, and region merging
[2].
C. Edge Based Image Segmentation
It is essential step for image segmentation process [3]. It
isolates an image into objects and its background. Edge
recognition separates the image by watching the adjustment
in power or pixels of a image. Gray histogram and Gradient
are two fundamental routines for edge detection for image
segmentations [4]. A few administrators are utilized by edge
detection technique, i.e., Classical edge locators, zero
crossing, Laplacian of Guassian (LoG)[5], and color edge
detectors. [6].
D. Fuzzy Theory Based Image Segmentation
This technique is used in arrange to analyze images, and
give correct in order from any image. Fuzzification function
can be utilized to expel noise from images too [7]. A gray
scale image can be effortlessly changed into a fuzzy image
by utilizing a fuzzification function. Diverse morphological
operations can be joined with fuzzy method to show signs of
improvement results [8]. Fuzzy k-Means and Fuzzy C-
means (FCM) are generally utilized techniques as a part of
image processing [9].
E. ANN Based Image Segmentation
In Artificial Neural Network, each neuron is equivalent to
the pixel of an image. Image is mapped to the neural
network. Image as neural network is prepared utilizing
preparing tests, and after that association between neurons,
i.e., pixels are found. At that point the new images are
segmented from the prepared image [10]. A portion of the
generally utilized neural network for images division are
Hopfield, BPNN, FFNN, MLFF, MLP, SOM, and PCNN.
Segmentation of images utilizing neural network is execute
as a part of two stages, i.e., pixel classification and edge
detection [11].
III. CLUSTERING ALGORITHMS
Clustering algorithms can be divided into three categories as
follows [12]
 Partitioning Clustering
 Hierarchical Clustering
 Density-based Clustering
Partitioning algorithms attempt to determine k clusters that
optimize a certain, often distance-based criterion function.
Hierarchical algorithms create a hierarchical decomposition
of the database that can be presented as a dendrogram.
Density-based algorithms search for dense regions in the
data space that are separated from one another by low
density noise regions.
A. Partitioning Clustering
Partitioning clustering partitions an arrangement of N items
into k clusters such that the segments streamline a sure rule
function. Every cluster is indicated to by the centroid of the
group, or by medoid, for instance k-medoids. Normally, k
seeds are arbitrarily chosen and afterward a relocation
scheme iteratively reassigns focuses between groups to
upgrade the clustering paradigm. The minimization of the
square-error basis - whole of squared Euclidean distances of
focuses from their nearest cluster centroid, is the most
usually utilized. A genuine downside of partitioning
algorithms is that there are various conceivable solutions.
B. K- Means
K-means is perhaps the most popular clustering method in
metric spaces. Initially k cluster centroids[13,12] are
selected at random; k-means then reassigns all the points to
their nearest centroids and recomputed centroids of the
newly assembled groups. The iterative relocation continues
until the criterion function, e.g. square-error converges.
Despite its wide popularity, k-means is very sensitive to
noise and outliers since a small number of such data can
substantially influence the centroids. Other weaknesses are
sensitivity to initialization, entrapments into local optima,
poor cluster descriptors, inability to deal with clusters of
arbitrary shape, size and density, reliance on user to specify
the number of clusters. Finally, this algorithm aims at
minimizing an objective function; in this case a squared
error function.
∑ ∑ ∑ ∑‖ ‖
where || xi(j) – cj ||2 is a chosen distance measure
between a data point xi(j)and the cluster centre Cj, is an
indicator of the distance of the n data points from their
respective cluster centres.
C. Fuzzy C- Means
In this algorithm, each point has a level of having a place
with groups, as in fuzzy algorithm, as opposed to having a
place altogether too simply single group. Consequently,
focuses on the edge of a group can be in the bunch to a
littler degree than focuses in the focal point of group. For
every point x we have a coefficient giving the level of being
in the kth bunch uk(x). More often than not, the total of
those coefficients is characterized to be
∑
With fuzzy k-means, the centroid of a cluster is the
mean of all points, weighted by their degree of belonging to
the cluster:
∑
∑
The degree of fit in is related to the inverse of the
distance to the cluster center:
then, the coefficients are normalized and fuzzy
fired with a real parameter m > 1 so that their sum is 1.
So
∑ ( )

For m equivalent to 2, this is equal to normalizing the
coefficient straightly to make their whole 1. At the point
when m is near 1, then group focus nearest to the fact of the
matter is given substantially more weight than the others,
and the calculation is like k-means. The fuzzy k-means
calculation is fundamentally the same to the kmeans
calculation:
1) Choose various bunches.
2) Assign randomly to every point coefficients for being in
the bunches.
3) Repeat until the calculation has united (that is, the
coefficients' change between two cycles is close to ɛ, the
given affect ability threshold)
D. Hierarchical Algorithms
Unlike partitioning methods that create a single partition,
hierarchical algorithms[5] produce a nested sequence (or
dendrogram) of clusters, with a single all-inclusive cluster at
the top and singleton clusters of individual points at the
bottom. The hierarchy can be formed in top-down (divisive)
or bottom-up (agglomerative) fashion and need not
necessarily be extended to the extremes. The merging or
splitting stops once the desired number of clusters has been
formed. Typically, each iteration involves merging or
splitting a pair of clusters based on a certain criterion, often
measuring the proximity between clusters.
E. Density-Based Clustering Algorithms
Density-based clustering methods assembly nearing objects
into clusters based on neighboring density conditions rather
than proximity connecting objects. These techniques see
clusters as thick regions being isolated by low density
uproarious regions. Density based methods have commotion
resilience, and can find non-curved clusters. Like various
leveled and apportioning strategies, density based methods
experience troubles in high dimensional spaces on account
of the inalienable shortage of the element space, which
thusly, diminishes any clustering tendency.
IV. VARIOUS OPTIMIZATION TECHNIQUES
Optimization is a usually encountered numerical problem in
all engineering disciplines. It actually means discovering the
best possible/desirable solution. Optimization problems are
far reaching and various, thus systems for taking care of
these issues should be a dynamic research topic.
Optimization algorithms can be either deterministic or
stochastic in nature. Previous techniques to take care of
enhancement issues require tremendous computational
efforts, which have a tendency to come up short as the issue
size increments. This is the motivation for employing bio
inspired stochastic optimization algorithms as
computationally efficient alternatives to deterministic
approach [14].
A. Evolutionary algorithms
EA‟s are most known, established algorithms among all
other Optimization algorithm. EA‟s use the methods used by
all living organisms to interact with each other. These
algorithms used this powerful strategy to find solution to
hard problems. EAs are non-deterministic algorithms or cost
based algorithms.
B. Swarm Intelligence based algorithm
This is based on collective behavior of organisms. SI works
on the implementation of groups of simple agents that are
based on the behavior of real world insect swarms, as a
problem solving tool.
1) Particle Swarm Optimization (PSO)
Particle Swarm Optimization (PSO) is a computational
insight arranged, stochastic, population based global
optimization technique proposed by Kennedy and Eberhart
in 1995[15]. It is motivated by the social activities of bird
flocking searching for food. PSO has been broadly
connected to numerous engineering optimization areas
because of its one of a kind searching method,
straightforward idea, computational proficiency, and simple
implementation.. In PSO, the term ―particlesɛ alludes to
population members which are mass-less and volume-less
(or with a discretionarily little mass or volume) and are
liable to speeds and increasing speeds towards a superior
method of conduct. Every particle in the swarm speaks to an
answer in a high-dimensional space with four vectors, its
current position, best position discovered in this way, the
best position found by its neighborhood so far and its
velocity and modifies its position in the hunt space in view
of the best position came to without anyone else's input
(pbest) and on the best position came to by its neighborhood
(gbest) during the process. In every iteration, each particle
updates its position and velocity.
2) Ant Colony Optimization:
In the normally, ants (firstly) wander randomly, and judging
food return for his or her colony while put down pheromone
trails. If next ants find this type of path, they're possible not
to keep traveling randomly, but to in its place track the trail,
returning and reinforcing it as long as they eventually get
food. After a while, however, the pheromone trail starts to
evaporate, thus reducing its attractive strength. The
additional time it requires for an ant to see down the trail
and rear, the additional time the pheromones need certainly
to evaporate. A fast path, in comparison, find marched over
more frequently, and therefore the pheromone concentration
becomes higher on shorter paths than longer ones.
Pheromone evaporation even offers the benefit of avoiding
the convergence to a locally optimal solution. If there were
no evaporation at all, the paths chosen by the original ants
would are often excessively appealing to the next ones.
Because case, the exploration of an ideal solution is space
could possibly be constrained. Thus, when one ant finds a
good (i.e., short) path from the colony to a food source,
other ants are susceptible to follow that path, and positive
feedback eventually leads to any or all or any the ants‟
carrying out a single path.
3) Artificial Bee Colony Algorithm (ABCA)
ABCA is based on the behaviour of the bee‟s environment.
No of swarm intelligence algorithms are present. These
algorithms are classified on the behavior, foraging
behaviour and mating behaviour. Examples of algorithms
simulating the foraging behaviour of the bees consist of the
Artificial Bee Colony (ABCA), the Virtual Bee algorithm
proposed by Yang, the Bee Swarm Optimization algorithm
proposed by Drias et al., the BeeHive algorithm proposed by
Wedde et al., and the Bee Colony Optimization algorithm
planned by Teodorovic and Dell„Orco . An individual entity
(e.g., a bee in a bee colony) exhibit a simple set of behaviour

policies (e.g., migration, replication, death), but a group of
entities (e.g., a bee colony) shows complex emergent
behaviour with useful properties such as scalability and
adaptability. Artificial Bee Colony is a predominant
algorithm simulating the intelligent foraging behaviour of a
honeybee swarm, proposed by Karaboga and Basturk [16].In
ABC algorithm, the colony of artificial bees contains three
groups of bees: employed bees, onlookers and scouts.
V. LITERATURE REVIEW
Waseem Khan[17] In this article, various techniques of
image segmentation has been discussed, an overview of all
related image segmentation techniques has been presented in
this paper. Recent research in image segmentation
techniques is presented in this paper. After the analysis of
different techniques of image segmentation, it is observed
that a hybrid solution for image segmentation consists of
two or more techniques is being the best approach to solve
the problem of image segmentation.
Abhay Sharmaet.al [18] Particle swarm
optimization is the nature motivated computational search
and optimization approach which was produced on the
premise of conduct of swarm. As of late every single field
of exploration is using the properties of PSO. One of the
prominent field of exploration is image segmentation which
is likewise quickest developing field. Taking the upsides of
joining PSO with diverse image segmentation strategy
numerous specialists has proposed different research papers
with improvement of different parameter. In this paper we
surveyed some paper and attempt to give late patterns and
methods included in image segmentation with PSO.
Fahd M. A. Mohsen et.al [19] In this work, PSO
has been used to produce a new optimization-based image
segmentation method, PSOTH. In the PSOTH method, the
algorithm of PSO tries to find a near optimal segmentation
for a given image using a fitness function. PSO is a flexible
optimization method, where many objective functions can
be used. For this reason, a new quantitative evaluation
function for segmented images has been proposed in this
paper. So in the PSOTH method, the new evaluation
function has been used as a fitness function for the
algorithm of PSO.
Shi Na et al. [10] conversed the standard k-means
clustering algorithm and analyzes the short-comings of
standard k-means algorithm,
Such as the k-means clustering algorithm compute
in each iteration, the distance between each data object and
all cluster centers, which create the efficiency of clustering,
was not high. This paper proposes an improved k-means
algorithm in order to solve this question, requiring a simple
data structure to store little information in each iteration,
which was to be used in the next iteration. The enhanced
method avoids computing the distance of each data object to
the cluster centers repeatedly, reduction the execution time.
Experimental results explain that the enhanced method can
efficiently get better the speed of clustering and
accurateness, dropping the computational complexity of the
k-means.
INA SINGH, Optimal Selection of initial clusters is
a challenging task in image segmentation. Wrongly selected
clusters lead to poor results. We have designed a technique
for image segmentation using K-means clustering and Ant
colony Optimization for Abdominal CT images to segment
the liver region. This technique is much better in segmenting
liver than the other models. Thus it is concluded that for
abdominal CT images, ACO based K-means clustering is
better than many other techniques [20].
VI. CONCLUSION
Image segmentation is a main step in the area of image
analysis and compression. The process of dividing an image
into multiple regions (set of pixels) is known as Image
segmentation. It is detected that there is no any suitable
method for image segmentation because the solution of
image segmentation is depends on several factors, i.e.,
texture, pixel color, intensity, matching of images, image
content, and problem domain. The paper shows the detailed
explanation about how the segmentation is done by using
each technique separately. This natural technique of
computing provides a number of ways for resolving the real
world problems, more efficiently and quickly with accuracy.
Among them, ACO has a wide range of applicability which
makes it one of the important and efficient techniques for its
wide range of applications.
REFERENCES
[1] K. K. Rahini S. S. Sudha “Review of Image
Segmentation Techniques: A Survey” International
Journal of Advanced Research in Computer Science
and Software Engineering Volume 4, Issue 7, July
2014.
[2] H. G. Kaganami and Z. Beij, “Region based detection
versus edge detection,” IEEE Transactions on
Intelligent Information Hiding and Multimedia Signal
Processing, pp. 1217-1221, 2009.
[3] M. Sarif, M. Raza, and S. Mohsin, “Face recognition
using edge information and DCT,” Sindh Univ. Res.
Jour. (Sci. Ser.), vol. 43, no. 2, pp. 209-214, 2011.
[4] S. Lakshmi and D. V. Sankaranarayanan, “A study of
edge detection techniques for segmentation computing
approaches,” IJCA Special Issue on “Computer Aided
Soft Computing Techniques for Imaging and
Biomedical Applications” CASCT, 2010.
[5] M Sharif, S Mohsin, M. Y. Javed, and M. A. Ali ,
“Single image face recognition using laplacian of
gaussian and discrete cosine transforms,” Int. Arab J.
Inf. Technol., vol. 9, no. 6, pp. 562-570, 2012.
[6] B. Sumengen and B. Manjunath, “Multi-scale edge
detection and image segmentation,” in Proc. European
Signal Processing Conference, 2005.
[7] S. Naz, H. Majeed, and H. Irshad, “Image
segmentation using fuzzy clustering: A survey,” in
Proc. 6th International Conference on Emerging
Technologies, 2010, pp. 181-186.
[8] I. Irum, M. Raza, and M. Sharif, “Morphological
techniques for medical images: A review,” Research
Journal of Applied Sciences, vol. 4, 2012.
[9] D. Hu and X. Tian, “A multi-directions algorithm for
edge detection based on fuzzy mathematical
morphology,” in Proc. 16th International Conference
on Artificial Reality and Telexistence--Workshops,
2006, pp. 361-364.

[10] B. J. Zwaag, K. Slump, and L. Spaanenburg, “Analysis
of neural networks for edge detection,” 2002.
[11] D. Suganthi and Dr. S. Purushothaman, “MRI
segmentation using echo state neural network,”
International Journal of Image Processing, vol. 2, no.
1, 2008
[12] Biswas, G., Weingberg, J. and Fisher, D.H.,
ITERATE: “A conceptual clustering algorithm for data
mining. IEEE Transactions on Systems, Man, and
Cybernetics”. v28C. Page(s):219-230.
[13] Hanning Chen , Yunlong Zhu, “Optimization based on
symbiotic multi-species coevolution; journal on
Applied Mathematics and Computation “ (2008)
[14] Dorigo, M., Maniezzo, V., & Colorni, A.. “Ant
System: Optimization by a colony of cooperating
agents. IEEE Transactions on Systems, Man, and
Cybernetics “– Part B, 26, (1996),Page(s): 29–41.
[15] Kennedy, J.; Eberhart, R. "Particle Swarm
Optimization". Proceedings of IEEE International
Conference on Neural Networks. (1995). Page(s):
1942–1948.
[16] D. Karaboga, B. Basturk, “A powerful and efficient
algorithm for numerical function optimization:
artificial bee colony (ABC) algorithm” Journal of
Global Optimization 39 (2007) Page(s): 459–471
[17] Waseem Khan “Image Segmentation Techniques: A
Survey” Journal of Image and Graphics Vol. 1, No. 4,
December 2013.
[18] Abhay Sharma, Rekha Chaturvedi, Dr. Umesh Kr.
Dwivedi” Recent Trends and Techniques in Image
Segmentation using Particle Swarm Optimization-a
Survey” International Journal of Scientific and
Research Publications, Volume 5, Issue 6, June 2015.
[19] Fahd M. A. Mohsen, Mohiy M. Hadhoud and Khalid
Amin “A new Optimization-Based Image
Segmentation method By Particle Swarm
Optimization” (IJACSA) International Journal of
Advanced Computer Science and Applications, Special
Issue on Image Processing and Analysis.
[20] Ina Singh, Neelakshi Gupta “Liver Image
Segmentation Using Ant Colony Based Kmeans
Clustering And Level Sets” Proceedings of 33th IRF
International Conference, 26th July 2015.

A Review on Image Segmentation using Clustering and Swarm Optimization Techniques

More Related Content

What's hot (20)

Viewers also liked (11)

Similar to A Review on Image Segmentation using Clustering and Swarm Optimization Techniques (20)

More from IJSRD (20)

Recently uploaded (20)

A Review on Image Segmentation using Clustering and Swarm Optimization Techniques