Global Journal of Science, Engineering and Technology (ISSN: 2322-2441)
Issue 14, 2013, pp. 132-141
© GJSET Publishing, 2013.
http://www.gjset.org
132
Color Image Segmentation Using a Weighted kernel-based Fuzzy C-
Means Algorithm
Siavash Alipour
1
, Mousa Nazari
2
, Jamshid Shanbehzadeh
3
1,2
student with Department of Computer Engineering Kharazmi University
1
Siavash.alipour@tmu.ac.ir ,
2
Nazari.mousa@gmail.com
3
Associate Professor with Department of Computer Engineering Kharazmi University,
jamside@tmu.ac.ir
Abstract: Color image segmentation plays an important role in computer vision and image processing applications.
Kernel-based fuzzy C-means (KFCM) is well known and powerful methods used in image segmentation.
Moreover, an appropriate assigning weight to features can improve its performance. This paper focuses on
improving the image segmentation capabilities of KFCM based on feature weighting. It employs Entropy concept
to measure the weight of features based on statistical variations viewpoint in KFCM. We compare the segmentation
results of the proposed method with the well know algorithms along the same line that used weight selection
procedure in FCM algorithm. Our simulation results reveal that the proposed algorithm provides greater
segmentation performance for color image segmentation according to cluster validity function.
Keywords: kernel-based fuzzy C-means; Entropy; FCM; weight selection; Variation; Color image segmentation
1. Introduction
Color image segmentation plays an important role in
computer vision and image processing applications.
Segmentation is to partition an image into regions
that are homogeneous according to a criterion, such
as intensity, color, tone, texture and, etc [1]and, it
is considered as one of the most difficult tasks in
image processing and affects the quality of the
final results of an image processing or machine
vision system. Generally, the gray level image
segmentation techniques can be divided into four
categories, thresholding, clustering, edge detection,
and region extraction [2]. At present, these
techniques can be extended to color images
expressed in different color spaces. These spaces are
obtained by using the linear and non-linear
transformations of the RGB color space. Many
different techniques of color image segmentation
have been developed and detailed in the literature [3].
Nowadays, the clustering methods [4, 5] are
considered as one type of fundamental tool for image
segmentation. In clustering methods where clustering
numbers are known in prior, fuzzy C-Means (FCM)
algorithm is one of the most widely used methods
due to its flexibility in clustering pixels of segmented