IRJET- Image Segmentation Techniques: A Survey

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 05 Issue: 04 | Apr-2018 www.irjet.net p-ISSN: 2395-0072
© 2018, IRJET | Impact Factor value: 6.171 | ISO 9001:2008 Certified Journal | Page 51
Image Segmentation Techniques: A Survey
Rahul Basak1, Surya Chakraborty2, Aditya Kumar Mondal3, Satarupa Bagchi Biswas4
12,3 Dept of Information Technology, Heritage Institute of Technology, Kolkata, West Bengal, India
4 Asst. Professor Dept of Information Technology, Heritage Institute of Technology, Kolkata, West Bengal, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Technology is always been a factor which
accelerate the time. Among the several hands of it, Image
Processing or better to say Digital Image processing is most
important portion to study. And Image Segmentation is one of
the hotspot of Digital Image Processing. There is no any
general solution available for this. Several general-purpose
techniques have been developed for Image Segmentation
process. This paper addresses some of the most important
techniques from the brunch and represents a survey on them.
Key Words: Image segmentation, Image Segmentation
Techniques, Image Processing, Histogram Technique,
K-means, Fuzzy C-means, Watershed Technique,
Clustering techniques.
1. INTRODUCTION
In Image Processing it’s been a always important part to
segment an image into multiple section, to study or evaluate
them properly. A image can be segment depends upon so
many factors like colours, Textures, grey scalevalue[21]etc.
and also there is no any general rule available for this. So,
naturally there is so many general-purposedapproachesare
available for the segmentation process, which leads to a
separate study on Image Segmentation. In Image
Segmentation, it means divides a image into multiple parts
[21] which are definable, actionable, profitable and
accessible. And those parts can also be evaluated separately
without interfere into each other. And as mentioned there
are so many techniques are available [20] in the world. But
according to their behaviours of segmentation process they
are divided into five methods as described in figure: 1.
Fig-1: Classification of Segmentation Technique
Here is some example of techniques for each method. For
Threshold Based: histogram based techniques, iterative
thresholding, otsu method, etc. For Edge Based: the Hough
transformation, watershed segmentation, snakes, etc. For
Region Based: region growing, region merging, split and
merge method etc. pyramid tree and scale method, texture
method, Fourier technique, co-occurrence matrices, etc. For
Cluster Based: agglomerative clustering, K-means, Fuzzy C-
means, etc. For Matching Based: templatematching,etc.This
paper address fourtechniquesfromabove whichare belongs
to different method. They are Histogram Based, K-means,
Fuzzy C-means (Both K-means and Fuzzy C-means belongs
to the same method.), Watershed.
If you look around into the world from your right next
mobile phone to any big industry, or any hospital you can
find Image Segmentation is a big deal over there. And some
of the cases it’s really need to be more accurate as it can be
cause security for a system for example your mobile phone
or personal computer can be open with finger print scan or
by face detection which is developed by imagesegmentation
process. And if you go in to medical science field you can
find it also very useful like locate a tumor, details study on
anatomical structure, measuring tissue volume those things
are done by Image Segmentation.Inotherfieldslikemachine
view in Industry, objectlocatingfromSatellite,contentbased
image retrieval. Keeping in the mind the importance of
Image Segmentation the developers always invented new
approaches sometime by merging the best from different
techniques or algorithms to make the required result as
accurate as possible.
2. SEGMENTATION TECHNIQUES
Now weare going to discuss those four particulartechniques
by some research paper available from recent study.
2.1. Histogram Based
It is one of the simplest way to receive segmentedpartsof an
image from the histogram of it. In the classic approach first,
the histogram of the image is made according to the color
and intensity value, the cluster can be define. Based on the
clusters got from the image, segmentation is done but the
drawback over here is we cannot get much required level of
details in the segmented pictures.
More technically, The histogram based techniques is
dependent on the success of the estimating the threshold
value that separates the two homogonous region [23] of
the object and background of an image. Histogram based
thresholding is applied to obtainall possibleuniformregions
in the image.
Salem Saleh Al-amri et al. [1] states that segmentation
process based on Histogram can be done in two way. Shown
in figure 2 [1]

Fig -2: Techniques classification of Histogram Technique
First category is an image should be partitionedwith respect
to some of its features like image intensity, edgesetc.Second
category is to partition an image into several regions that
can clearly segregate an image into several regions that may
be with respect to intensity or colour value.
Histogram Threshold approach belongs to intensity, regions
of an image; i.e., with respect to colour intensity or with
respect to colour regions, a corresponding Histogram is
placed for that image which denotes;theamounteveryno.of
pixels that are presented in that image. Threshold
segmentation techniques can be classified into three
different classes:
 Taking in account the local properties of an image
like its pixel value, neighbourhood pixel details etc.
 Global techniques segment an image on the basis of
information obtain globally (By using image
histogram; global texture properties).
 Splitting, merging and growing techniques for an
image( use both the notions of homogeneity and
geometrical proximity)toobtaingoodsegmentation
results for an image.
They shown [1] the comparative studiesappliedbyusingfive
techniquesofthresholdsegmentationtechniques.Andamong
them HDT become one of best process. In this paper, the
researchers have worked on a particular domain of input
data. But there may be a chance that the result will change,
with other input data. The approach in [2] theyhaveusedthe
thresholding by choosing the mean or medianvalueofapixel
which is the key parameter in thresholding.Ifanyobjectpixel
is brighter than the background, that should also posses
higher value than the average. In a noiseless image having
uniform background and object values, the mean or median
will work well as the threshold as there the segregation of
object from background is easier,however,thiswillgenerally
not be the case. A more complicated approach is to create a
histogram of the image by calculating the intensity values of
pixels and calculate the threshold value by taking the valley
point of the placed histogram. The histogram approach like
that there exists some average value for the background
pixels and object pixels, although the actual pixelvalueshave
some variation around these average values. They have also
enlightenedonAdaptiveThresholding:differentthresholding
is applied on spatial variation of pixel's intensity for a given
image. In [3] they have applied Histogram technique along
with Fuzzy CMeanstechnique.Aclusteringbasedapproachis
the segregation of objects into similar groups, or more
precisely, the partitioning ofadatasetintosubsets(clusters),
so the data in the set can share common clusters. . Many
clustering schemes are categorized based on their special
characteristic, like the hard clustering scheme and the fuzzy
clustering scheme. One of the most popular clustering
methodsusedinimagesegmentationisFuzzyC-means(FCM)
algorithm because it can detectthedegreeofparticipationfor
an pixel and can hold much more information about pixel
details. Although conventional FCM is not associated with
spatial context information and it is sensitive to noise and
imaging artefacts. The fuzzy clusteringalgorithmFCMisthen
employed in the proposed approach to achieve proper
segmentation. To prevent & detect noise, the spatial
probability of neighbouring pixels is combined with the
conventional FCM. By using an efficient algorithm which can
effectively removenoise,theinputnoisymedicalimagefirstly
the noise is removed so that it can improve its robustness
further. When the spatial information combines with
traditional FCM it is clear that it will take longer time to
converge andalso there existslotsofpossibilitiestoconverge
in the local minima. Thus, in the presented approach, for
avoiding local minima, the parameters of the FCM algorithm
are initialized using histogram. Generally, clustering is to
segregate an image into different clusters with the intensity
values(& colour values too) of pixels but it does not bother
about the spatial information of the respective pixel. Hence
the histogram based FCM converges very quickly in
comparison with conventional FCM. In thisway[4]theyhave
discussed about two more recent and fruitful segmentation
process which is termed as class drivensegmentation;where
object class models liable to propose objectlocalizationsthat
is efficient in image segmentation. Another method is
interactive segmentation; in which user gives approximate
segmentations and then refines and gather the auto created
image based segmentations into groups. Here it is discussed
to show that if an appropriate distance measure is used;
equal or superior recognition results can be obtained by a
single class model, and also to explain why this result comes
about. Moreover this paper enlighten on to show that pixel-
wise segmentations can be obtained from sliding windows
using class models. In addition, an object category is
represented by a single histogram of dense visualwords,and
then look for the effectiveness of this representation for
segmentation. Actually the advantage of a single class
histogram it can be represented efficiently in terms of
computation.
2.2. Clustering Method
Clustering is a significant task in data analysis and data
mining applications (Which can be applied on Image
segmentation process). It is the task of arrangement a set of
objects so that objects in the identical group are more
related to each other than to those in other groups(clusters)

A good clustering method will produce [23] high superiority
clusters with high intra-class similarity and low inter-class
similarity. Clustering algorithms can be categorized into
partition-based algorithms, hierarchical-based algorithms,
density-based algorithms and grid-based algorithms.
Partitioning clustering algorithm splitsthedata pointsintok
partition, where each partition represents a cluster.
Fig -3: Clustering process with FCM and K Methods
We are discussing two of the most important Clustering
Techniques here. They are:
2.2.1. K-Means Clustering
The K-means techniqueisusedtopartitionanimage
into K clusters. Every pixel only belongs to a particular
cluster. The clustering is done based on either pixel colour
intensity of the image or texture, location, or a combination
of these factors. The K value can be selected manually,
randomly, or by a heuristic approach. Although this
algorithm is guaranteed to converge, but it is hard to say
about the optimality of solution. The quality of the result
depends on the initial set of clusters and the selection of K
value. Also, K-means clustering is not suitable fortheimages
that have fuzziness. More technically It is a partitionmethod
technique which finds mutual exclusive clusters ofspherical
shape. It generates a specific number of disjoint, flat(non-
hierarchical) [24] clusters. Stastical method can be used to
cluster to assign rank values to the cluster categorical
data.
In this paper [5] they have discussed about Subtractive
clustering method which is an efficient technique that can
find the optimal data point, which serves as clustercentroid;
based on the density of neighbourhood data points. It
estimates the number and initial location of the cluster
centres. It distributes the data space into gridding point and
then calculates the distance of every data point from actual
data point. So the grid point with many data point nearby
possess high potential value; so this grid point with highest
potential value will be choose as first cluster centre. After
selecting the first cluster centre the second cluster centre to
be found by calculating the highest potential value in the
remaining grid points. This method of acquiring new cluster
centre and reducing the potential of surrounding grid point.
This process will be circulated till the convergence of every
grid point; that their potential falls under pre determined
threshold value. One problem is that, as the data dimension
is increased; corresponding computation complexity is also
increased exponentially.
Sadia Basar et al. [6] they has proposed system which will
create the temporary and individual clusters that helps to
find the optimal threshold value. The optimal cluster value
can be calculated by applying K-Means clustering algorithm.
They have also used Histogram for each individual colour
domain In RGB domain to find the individual peak, the
relative distance of each are also considered to calculatethe
mean value of each cluster for final segmentation. As per
their work, an image is firstly loaded, then it will be created
the temporary and individual clusters that will find the
optimal threshold value. K-means clustering algorithm is
used to find the optimal cluster value calculationandfeature
has been extracted i.e. pixel colourvalue,pixel intensitylevel
and regions. After the feature extraction, for RGB domain,
three histograms will be generated corresponding to Red,
Green, and Blue. Then peaks are identified for Red, Blue and
Green and then relative distance from peak to every data
point is calculated. After calculating the relative distance ,
the mean value should be calculated for each cluster; the
image is segmented by using the mean cluster value, finally.
Hong Liu et al. [7] they have focused on Content-Based
Image Retrieval (CBIR) technique which correlates on
searching for an image in database as if to the query for the
image, according to the image features related to content. In
CBIR technique, feature vectors are extracted from images
which exist in a very high-dimensional space. This high
dimensionality causes for high computational complexity in
calculation for similarity retrieval, and become inefficient in
case of indexing or searching. CIBR integrates semantic
cluster classifier with k-means algorithm. These also
improve the efficiency. This paper has proposed to
incorporate a clustering component in the region-based
image retrieval that reduces the inefficiencyofsearchingthe
whole document; the purpose can be served by searching
only the clusters that are nearer to the query target.Here[8]
they have mainly discussed about Intrusion and three
intrusion datasets viz KDDCup99, NSLKDD, and GureKDD
are implemented with the help of K-Mean clustering
technique. Intrusion is actually attempting to compromise
the confidentiality, integrity and availability of an
information resource. Intrusion detection istheprocessthat
monitors the network or system activities from any
corrupting or from any malicious. It automates the process

and counteract the intrusive efforts and intrusiveeffortscan
be effected by insiders (like any system failure) or outsiders
(any malicious attack) in the system. The pre-processed
datasets are furthernormalized(toremoveanyredundancy)
by applying various data pre-processing techniques, and
then applied as input to the models. Ramaraj.M et al. In their
study [9] a new approach has been discussing of brain
tumour the image, by applying K-means algorithm along
with Fuzzy c-means technique.. The statistical parameters
for a tumour are calculated by applying K-means & Fuzzy C
method. They have incorporated K-Means algorithm with
FCM in association with Hierarchical clustering; which
proper way to find the distance that is applied in traditional
clustering method. That method also generates a tree
structure (or a dendrograms) and it’s distance matrix.
Bottom-up & Top-down approaches are available for
Hierarchical clustering.
2.2.2. Fuzzy-C means Clustering
Fuzzy clustering (or Soft Clustering) is a techniqueforimage
segmentation in which each data point can belong to more
than one cluster or partition. Membership grades are
assigned to each of the data points. These membership
grades indicate the degree of participation that indicates
which data point belongs to which cluster. Thus, points on
the edge of a cluster, with lower membership value, indicate
that the cluster to a lesser degree of participation than
central cluster point. Technically The use of fuzzy set
provides imprecise class membership function. of the key
constituents of soft computing in handling challenges
posed by massive collections of natural data. The central
idea [25] in fuzzy clustering is the non-unique
partitioning of the data into a collection of clusters. The
data points are assigned membership values for each of the
clusters and fuzzy clustering algorithm allow the clusters to
grow into their natural shapes. In this case [10] they classify
image using statistical features (mean and standard
deviation of pixel colour values) which is a simple but
powerful method for text as well as image segmentation. A
systematic structure is followed by these features which
leads that segregation one from another. Theyidentifiedthis
segregation in the form of class clustering; Fuzzy C-Means
method which is used to determine each cluster location;
using this technique, maximum membership defuzzification
and neighbourhood smoothing is achieved. The steps that
they have used are demonstrated below:
Fig- 4: Process described [10] for FCM method
After achieving stability of the transient iterative mapping,
all the pixel blocks are considered that they belong to one of
the predetermined regions-this is Defuzzification[10]. And
they have applied smoothing, which is the technique that
reduces noise retaining the boundary object. To avoid
blurring effects the output is calculated using pixel values
from the same cluster. Segmentation and classification
becomes difficult to handle when any multiplicative noise is
appeared in Synthetic Aperture Radar (SAR) images.
Although by help of a Fuzzy C-means (FCM) algorithm and
its variants; satisfactory segmentation results can be
achieved and they are robust to noises. This letter [11]
presents a kernel FCM algorithm where pixel intensity and
location information are enlighten for SAR image
segmentation. They incorporate a weighted fuzzyfactorinto
the objective function, which works as intensity distancesof
all neighbouring pixels simultaneously. By the help of this
which is worth emphasizing that the spatial distance is not
sufficient to reflect the relationship between the neighbour
pixel and the central pixel. By which they have segmented
the image. In this paper [12] they have applied a developed
algorithm to the segmentation and classification of Multi-
colour Fluorescence In Situ Hybridization (M-FISH) images,
& this kind of images can be used to detect chromosomal
abnormalities for cancer detection and any genetic disease.
By introducing a gain field, this algorithm enhanced the
general fuzzy c-means (FCM) clustering algorithm that
models and corrects the pixel intensity in homogeneities;be
affected by microscope imaging system. The gain field
regulates intensity cluster centre that reduces the error;
without affecting the homogeneously distributedintensities

over the image. Long Chen et al. [13] introduced a
generalized multiple-kernel fuzzy C-means (FCM)(MKFCM)
methodology;asa framework forimage-segmentation.In the
framework, the composite kernels are used in the kernel
FCM (KFCM), is just a integrating form of multiple kernels.
The proposed MKFCM algorithm provides flexibility to fuse
different pixel informationin image-segmentationproblems.
The kernel FCM (KFCM) algorithm is an extension of FCM,
which connects the original with Hilbert space by some
transform function. When this mapping is done, the data are
more easily to be separated or clustered. It can be defined
different kernel functions purposely for the intensity
information, the texture information and the combinationof
these kernel functions are applied the composite kernel in
MKFCM (including LMKFCM) to obtain better image-
segmentation results.
Yannis A. Tolias [14] has incorporated spatial constraints
into the results of conventional Fuzzy clustering technique
for solving image segmentation problems. They proposedof
imposing spatial constraints is based on a voting scheme
over a neighbourhood that is evaluated on a cluster basis.
The basic criterion for imposing spatial constraints over a
neighbourhood is ; when we deal with a homogeneous
region, either of low or high membership to a cluster, the
fuzzy partition matrix should be updated in such a way that
describes the membership of the majority of the pixel
neighbours to the cluster.
2.3. Watershed Transformation
The idea of watershed transform [26] is straightforward by
the intuition from geography. The main goal of watershed
segmentation algorithm is to find the “watershedlines”in an
image in order to separate the distinct regions. To imagine
the pixel values of an image is a 3D topographic chart,where
x and y denote the coordinate of plane, and z denotes the
pixel value. The algorithm starts to pour water in the
topographic chart from the lowest basin to the highest peak.
In the process, we may detect some peaks disjoined the
catchment basins, called as “dam”. The watershed algorithm
is one of the most powerful morphological tools for image
segmentation.
Fig-5: Catchment basin and Watershed line. [26]
They have fold the paper [15] in two. In first they have
present a critical review of several definition of Watershed
Transformation and associated sequential algorithm.Andin
the second main current approaches towards parallel
implementation of Watershed model, Depends upon
strategies, distinguishing between distributed memory and
shared memory architecture. They have also divided there
paper [16] in to two section. First, they define basictools,the
watershed transform. And then they show that this
transformation can be built by implementing a flooding
process on a grey-tone image. Using elementary
morphological operations like a geodesic skeleton and
reconstruction; this flooding process can be performed. By
applying this methodology, image segmentation operations
is discussed over here. Due to the application of Watershed
algorithm on a particular image to transform in gradient
image; causes a over segmentation. This leads, in the to the
part, to the introduction of a general methodology for
segmentation. They have enlightened on a transformation
viz. Homotopy modification. This complex tool is defined in
detail and various types of implementation are shownthere.
They add a new approach in there paper [17] approach of
Watershed Algorithm using DistanceTransformisappliedto
Image Segmentation. It is very common that withwatershed
transformation segmentation outcome image (segmented)
comes with a max level of noise; here to reduce that, they
have used Laplacian of Gaussian (LoG) edge detector
technique with the classic approach of watershed
transformation. By the help of this approach,asshowninthe
paper the result comes with the lesservalueof noises.Which
means subject visibility becomes cleareraftersegmentation.
Lamia J aafar et al.[18] they have also works with the same
issue that is to reduce over segmentation problem with a
new approach which is based on mathematical morphology.
More preciously they propose to adapt the topological
gradient method with the classical approach of Watershed
transformation technique. They have also illustrated the
numerical tests obtained from the result to show the
efficiency. In this paper [19] they have mainly concentrate
on the medical images issues. And based on this they have
developed theapproach.Magneticresonanceimage basically
comes with a much level of noise. Here they have targeted to
reduce those by the help of some marker techniques. They
propose to introduce the use of a previous probability
calculation in the watershed technique. Furthermore, they
introduce a method through the use of markers to combine;
the watershed transform and atlas registration,. And as the
result the images comes with less drawback ofsegmentation
like over-segmentation, sensitivity to noise or low signal to
noise ratio structures.
3. CONCLUSION
In this survey paper we have discussed several important
techniques for Image Segmentation.Someofrecentresearch
work on those techniques is also presentedanddiscussedby
us. After observing those techniques separately, we come to
the conclusion that, a hybrid solution for image
segmentation consists oftwo ormoretechniquesisbeingthe

best approach to solve the problem of image segmentation.
And that is what actually done by the most of the review
papers, In addition some of the authors also introduce some
mathematical approaches and edge detection techniques to
get the desired output. As there is no any universally
accepted method for image segmentation technique and
there is so many factors which create affect on the result.
Like: homogeneity of images, spatial characteristics of the
image, continuity, texture. Considering all above factors,
image segmentation remains a challenging problem in
image processing and computer vision.
ACKNOWLEDGEMENT
We would love to thank Professor Joydev Hazra for his
expert advice and encouragement throughout this project.
REFERENCES
[1] Salem Saleh Al-mria, N.V. Kalyankar and Khamitkar,
“Image Segmentation by Using Threshold Technique”,
JOURNAL OF COMPUTING, VOLUME 2, ISSUE 5, MAY
2010, ISSN 2151-9617,pp.83-86
[2] P.DanielRatnaRaju, G.Neelima, “Image Segmentation by
using Histogram Thresholding”, IJCSET |January 2012|
Vol 2, Issue 1, pp. 776-779
[3] Meenakshi M. Devikar and Mahesh Kumar
Jha,”SEGMENTATION OF IMAGES USING HISTOGRAM
BASED FCM CLUSTERING ALGORITHM AND SPATIAL
PROBABILITY”, International Journal of Advances in
Engineering & Technology, Mar. 2013,pp.-225-231
[4] F. Schroff, A. Criminisi, A. Zisserman, ”Single-Histogram
class models for Image Segmentation”, Indian
Conference on Computer Vision, Graphics and Image
Processing, 2006
[5] Nameirakpam Dhanachandra, Khumanthem Manglem
and Yambem Jina Chanu, “Image Segmentation using K-
means Clustering Algorithm and Subtractive Clustering
Algorithm”, Eleventh International Multi-Conferenceon
InformationProcessing-2015(IMCIP-2015)pp.764-771
[6] Sadia Basar, Awais Adnan, Nalia Habib Khan, Shahab
Haider, “COLOR IMAGE SEGMENTATION USING K-
MEANS CLASSIFICATION ON RGB HISTOGRAM”,Recent
Advances In Telecommunications, Informatics And
Educational Technologies, ISBN: 978-1-61804-262-0,
pp. -257-262
[7] Hong Liu, and Xiaohong Yu, “Application Research of k-
means Clustering Algorithm in Image Retrieval System”
Proceeding of the Second Symposium International
Computer Science and Computational
Technology(ISCSCT ‘09), China, 26-28,Dec. 2009
,pp.274-277
[8] Santosh Kumar Sahu and Sanjay KumarJena,“AStudyof
K-Means and C-Means Clustering Algorithms for
Intrusion Detection Product Development”,
International Journal of Innovation, Management and
Technology, Vol. 5, No. 3, June 14,pp. 207-213
[9] Ramaraj.M Dr. Antony Selvadoss Thanamani, ”IMAGE
SEGMENTATION BY USING K-MEANS CLUSTERING
ALGORITHM “, “Detection of Brain Tumor by Mining
MRI Images”IJARCCE,vol 2,issue 4,January 2013.
[10]SompornChuai-aree, ChidchanokLursinsap, Peraphon
Sophatsathit, and SuchadaSiripant,” Fuzzy C-Mean: A
Statistical Feature Classification of Text and Image
Segmentation Method”,AdvancedVirtual andIntelligent
Computing Center (AVIC) Department of Mathematics
Chulalongkorn University Bangkok, 10330,
Thailand,Page 1-6
[11]Deliang Xiang, Student Member, IEEE, Tao Tang, Canbin
Hu, Yu Li, and Yi Su, Senior Member, IEEE,“A Kernel
Clustering Algorithm With Fuzzy Factor: Application to
SAR Image Segmentation” IEEE GEOSCIENCE AND
REMOTE SENSING LETTERS, VOL. 11, NO. 7, JULY
2014,Page 1290-1294
[12]Hongbao Cao, Hong-Wen Deng, and Yu-Ping Wang,
Senior Member, IEEE,” Segmentation of M-FISH Images
for Improved Classification of Chromosomes With an
Adaptive Fuzzy C-means Clustering Algorithm”, EEE
TRANSACTIONS ON FUZZY SYSTEMS, VOL. 20, NO. 1,
FEBRUARY 2012,Page 1-7
[13]Long Chen, C. L. Philip Chen, “A Multiple-Kernel FuzzyC-
Means AlgorithmforImageSegmentation”,Fellow,IEEE,
and Mingzhu Lu, Student Member, IEEE,Page 1-12
[14]Yannis A. Tolias, Student Member, IEEE, and Stavros M.
Panas, Member, IEEE,” On Applying Spatial Constraints
in Fuzzy Image Clustering Using a Fuzzy Rule-Based
System”, IEEE SIGNAL PROCESSING LETTERS, VOL. 5,
NO. 10, OCTOBER 1998,Page 359-369
[15]Jos B.T.M. Roerdink and Arnold Meijster, “The
Watershed Transformation: Definatioin, Algorithmsand
Parallelization “, FundamentaInformaticae 41 (2001)
187–228, Page No: 1-40
[16]S.Beucher, ”The Watershed Transformation Applied to
Image Segmentation” Centre de
MorphologieMathématiqueEcole des Mines de Paris35,
rue Saint-Honoré 77305 FONTAINEBLEAU CEDEX
(France). Page No: 1-26
[17]P.P. Acharjya, A.sinha, S. Sarker, S. Dey, S. Ghosh, “A New
Approch of Watershed Algorithm Using Distance
Transform Applied to Image Segmentation.”,
International Journal of Innovative Research in
Computer and CommunicationEngineering,Vol.1,Issue
2, April 2013 Page No: 185-189

[18]Lamia J aafar Belaid and WalidMourou, “Image
Segmentation:AWatershedTransformationAlgorithm”,
Image Anal Stereol 2009;28:93-102, Page No: 93-102
[19]V. Grau, A. U. J. Mewes, M. Alcaniz, R. Kikinis, and S. K.
Warfield, “Improved Watershed Transform for Medical
Image Segmentation Using Prior Information.” IEEE
TRANSACTIONS ON MEDICAL IMAGING, VOL. 23, NO.4,
APRIL 2004 Page No: 447-458
[20]Nida M. Zaitouna, Musbah J. Aqelb, “Survey on Image
Segmentation Techniques” International Conferenceon
Communication, Management and Information
Technology (ICCMIT 2015), Procedia Computer Science
65 ( 2015 ) 797 – 806
[21]Waseem Khan, “Image Segmentation Techniques: A
Survey”, Journal of Image and Graphics Vol. 1, No. 4,
December 2013
[22]Amandeep Kaur Mann & Navneet Kaur, “Review Paper
on Clustering Techniques”, Global Journal of Computer
Science and Technology Software & Data Engineering”
Volume 13 Issue 5 Version 1.0 Year2013
[23]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, Page: 842-
845
[24]Aastha Joshi, Rajneet Kaur, “A Review: Comparative
Study of Various Clusterering Techniques in Data
Mining”, International Journal of Advanced Research in
Computer Science and Software Engineering, Volume 3,
Issue 3, March 2013, Page 55-57
[25]R.Suganya, R.Shanthi, “Algorithm-A Review Fuzzy C-
Means” International Journal of Scientific and Research
Publications, Volume 2, Issue 11 , November 2012 ISSN
2250-3153, Page: 1-3
[26]Farheen K. Siddiqui Prof.Vineet Richhariya,“AnEfficient
Image Segmentation Approach through Enhanced
Watershed Algorithm” International Conference on
Recent Trends in Applied Sciences with Engineering
Applications, Vol.4, No.6, 2013, Page no 1-8

IRJET- Image Segmentation Techniques: A Survey

More Related Content

What's hot (15)

Similar to IRJET- Image Segmentation Techniques: A Survey (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET- Image Segmentation Techniques: A Survey