Color Image Segmentation Based on a Modified K-means
Algorithm
Qi Zhang
Beijing Key Laboratory of
Information Service
Engineering
Union University
Beijing,China 100101
15032523829@163.com
Yue Chi
Beijing Key Laboratory of
Information Service
Engineering
Union University
Beijing,China 100101
Bornthisway209@126.com
Ning He
Beijing Key Laboratory of
Information Service
Engineering
Union University
Beijing,China 100101
xxthening@buu.edu.cn
ABSTRACT
Image segmentation is the first job in the process of machine
vision,It plays a key role for the later results of image analysis
results later,Therein color image segmentation is a more difficult
task.This paper proposed an improved bisecting k-means
algorithm.The image is segmented in LAB color space.This
method does not need to determine the value of K in
advance,Thus it improves the adaptability of the algorithm and
lowers the subjectivity of segmentation.The experiment has
shown that the algorithm can automatically determine the value of
K more accurately.
Categories and Subject Descriptors
I.4.6 [Segmentation]:Image Processing and Computer Vision-
Pixel classification,clustering.
General Terms
Algorithms, Performance, Design, Reliability, Experimentation,
Theory, Verification.
Keywords
Bisecting K-means,color image segmentation,L*a*b color space,
clustering
1. INTRODUCTION
So far,there have been a number of methods in image
segmentation field.As a relatively simple and efficient algorithm,
K-means clustering has been widely used.K-means algorithm
achieves a minimum gap in a clustering and the maximum
distance between clustering in unsupervised clustering
algorithm,So it has a better segmentation result.
The traditional clustering algorithms includes K-Means
algorithm,[1]Fuzzy C-Means algorithm , Moving K-Means
algorithm, Adaptive Fuzzy K-Means algorithm etc.K-Means
algorithm is the basic. Current research is tend to the optimization
of K-Means algorithm, and the selection of cluster number and
center, The mathematical model of K-means algorithm is the
process of calculating an optimal solution.Fuzzy C means
algorithm is a soft segmentation algorithm,and it is presented on
the basis of fuzzy set theory and K-means algorithm,Its
convergence has been proven in the literature [2],which has
proved that the FCM algorithm converges to a extreum.FCM
algorithm allows data to become a member of more than one
cluster through iteratively optimized objective function[3],The
greatest contribution of the algorithm is introducing the fuzzy
concept to the membership degree of image pixels.Therefore,the
FCM algorithm can retain more original image information than
original k-means algorithm.Salamah s. a. and others put forward
A new clustering algorithm.The new algorithm is Adaptive Fuzzy
- K - means (AFKM) clustering algorithm. The algorithm is based
on the concept of fuzziness and belonging which provides a better
and stronger adaptive clustering algorithm。[4] AFKM algorithm
can split the different types of images and the different
performance segmented regions。
The conventional clustering algorithms (such as the KM, FCM,
and clustering algorithms) have the following common
limitations:[5]
(1) The number of clusters for segmentation needs to be
defined.Different user defines different values of K, which leads
to subjective segmentation
(2) Clustering is processed on the base of local parameter (the
intensity value of pixel ) which has ignored the influence of
related spatial information to pixel parameter, and causes
information loss while segmenting.
This paper advanced a bisecting K-Means Clustering
Algorithm.The algorithm need not to determine the value of the K
in advance, at the same time, each segmentation needs only
determine two centroids.So reducing the requirements for initial
centroids.[6]
2. PROPOSED MODEL
This paper proposes a bisecting K-Means Clustering Algorithm.
Firstly, the image is divided into two categories. Then calculate
the information of the two part of images.Next, the obtained
result is compared with a threshold,according to the results of the
comparison to determine whether the class should be bisecting K-
Means segmentation.Repeat the above steps,until each category
cannot be split up again.
Algorithm flow chart is as follows:
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that
copies bear this notice and the full citation on the first page. Copyrights
for components of this work owned by others than ACM must be
honored. Abstracting with credit is permitted. To copy otherwise, or
republish, to post on servers or to redistribute to lists, requires prior
specific permission and/or a fee. Request permissions
from Permissions@acm.org.
ICIMCS '15, August 19-21, 2015, Zhangjiajie, Hunan, China
© 2015 ACM. ISBN 978-1-4503-3528-7/15/08…$15.00
DOI: http://dx.doi.org/10.1145/2808492.2808538