Anda di halaman 1dari 4

www.ijcait.

com

International Journal of Computer Applications & Information Technology Vol. 3, Issue II Aug-September 2013 (ISSN: 2278-7720)

Comparative Research on Fuzzy C-Means and K-Means Clustering Segmentation


Mamta Bharadwaj
Ambedkar Institute of Advanced Communication Technologies and Research, New Delhi 132-F (Top Floor), DDA Flats, Pocket -1,Mayur Vihar Phase-1, New Delhi-110091

Ankita Walia
Ambedkar Institute of Advanced Communication Technologies and Research, New Delhi 17-B/26, D.B.Gupta Road, Karol Bagh, New Delhi-110085

Hemant Tulsani
Ambedkar Institute of Advanced Communication Technologies and Research, New Delhi BC 14 C (EAST), Shalimar Bagh, New Delhi 1100

ABSTRACT
Image Processing is one of the most growing research areas. There are so many applications like remote sensing, image enhancement, weather forecasting, atmospheric study, medical, academics and radars etc. In this paper we will emphasize only on medical application by using some segmentation techniques. There are so many algorithms and techniques for image segmentation. This paper presents the comparative research of unsupervised segmentation-K-means clustering (Hard clustering) and Fuzzy C-means clustering(Soft clustering).these both techniques are comparative in terms of time. These algorithms are applied and performance is evaluated on the basis of clustering output. Fuzzy C-means produces close result to K-means clustering but it still requires more computation time than K-means clustering.

often produce missing edges or extra edges or extra edges on complex and noise images.

2)

Thresholding based methods:-Thresholding is the operation converting a multilevel image into a binary image i.e. it assigns the value of 0(background) or 1(objects) to each pixel of an image based on a comparison with some threshold value T (intensity or color value). When T is constant the approach is called global thresholding. Threshold selection is typically done iteratively however it is possible to derive automatic threshold selection algorithm. 3) Region based methods: - The goal of region based segmentation is to use image characteristics to map individual pixels in an input image to set of pixels called regions that might correspond to an object or a meaningful part of one. The various techniques are local techniques. Global techniques are splitting and merging. The performance of this algorithm depends on the application area and input image. 4) Graph based method: - In graph based methods segments are connected sub graphs obeying appropriate similarity criteria. The aim of image segmentation is to partition the nodes into a complete disjoint system where each partition is connected. Initially pixel forms a separate segment. The algorithm first constructs a grid graph and calculates initial weights. In the following steps-it chooses the edge with maximum weight and contracts its two end nodes. At the level of image this means that the segmentation on the two sides of the edges are merged. If the contracted nodes have common neighbors only one edge is kept and its weight is recalculated.
5) Cluster based methods: - In this technique data set is replaced by cluster pixels may belong together because of the some color , texture etc. K-means ,C-means and hierarchical clustering are the classes of cluster based methods. These are iterative technique that is used to partition an image into clusters that minimizes the variance between the pixel and the cluster center. This algorithm is guaranteed to converge. The quality of solution depends on the initial set of clusters. Clustering based method is further divided into three classes 1) Hierarchical clustering 2) Overlapping(fuzzy c-means) clustering 3) Exclusive(k-means) clustering Fuzzy C-means and K-means clustering are heart of this paper. Hierarchical clustering is based on union between two nearest clusters and progress until final result (few desired

Keywords
Image segmentation, hierarchical clustering, K-means clustering, fuzzy C-means clustering, image enhancement.

1. INTRODUCTION
Image segmentation one of the most important operation in image analysis. It is defined as the process of portioning a digital image into multiple regions (set of pixels). These pixels are similar with respect to some characteristic or computed property such as color, intensity or texture. Image segmentation is used in various applications and for all applications a single method cannot produce the desired result because images have different properties and some other factors like noise, brightness, color, shape and size etc. Most of the image regimentation algorithms are based on two characters of pixels gray level:DISCONTUITY:-The approach is based on portion an image based on abrupt change in intensity.eg. Gray level discontinuities points, lines, edges. 2) SIMILARITY:-The approach is based on portioning an image into regions that are similar according to a set of predefined criteria. There are five main categories of image segmentation 1)

1)

Edge detection methods:-Edge detection method

locates the pixels in the Image that corresponds to the edges of the objects seen in the image. The result is binary image with the detected edge pixels. Common algorithms used are sobel, prewitt and laplacian operators. These algorithms are suitable for images that are simple and noise free and will

P a g e | 44

www.ijcait.com

International Journal of Computer Applications & Information Technology Vol. 3, Issue II Aug-September 2013 (ISSN: 2278-7720)

clusters) reached. This is used in image and video retrieval. It can achieve 90% retrieval accuracy.

2. K-MEANS ALGORITHM

CLUSTERING

between each data points and the cluster nearest to it is computed and recorded in a data structure is suggested .Thus on the following iterations the distance between each data point and its previous nearest cluster is recomputed.

K-means clustering algorithm is an unsupervised method. It is used because it is simple and has relatively low computational complexity. It is suitable for biomedical image segmentation as the number of clusters (k) is usually known for images of particular regions of human anatomy. STEPS:1) Pick k-cluster center either randomly or based on some heuristic. 2) Assign each pixel in the image to the clusters that minimize the distance between pixel and cluster center. 3) Re-compute the cluster centers by averaging all of the pixels in the cluster. 4) Repeat the steps 2 and 3 until convergence is attained (no pixel change cluster). Finally this algorithm aims at minimizing an objective function
k n 2 ( j)

3. FUZZY C-MEANS ALGORITHM

CLUSTERING

Fuzzy C-means segmentation is over hard segmentation method because they could retain much more information from the original image. Fuzzy C-means (FCM) is a method of clustering which allows one piece of data to belong to two or more clusters. It works by assigning membership to each data point corresponding to each cluster center on the basis of distance between the cluster center and data point. More the data is near to the cluster center more is its membership towards the cluster center. Clearly summation of membership of each data point should be equal to one after each iteration membership and cluster are updated according to formula

1
ij c k 1
n ij i 1 n ij i 1

dij dik
m

2 1) m

J
j 1 i 1

|| xi

c j ||

xi
m

Vj

1, 2,3.......C .

xi ( j ) - Data points, c j
| xi ( j ) c j ||2

- cluster center

- chosen distance measure between data

n Number of data points.

points and cluster center. n- Data points, k-cluster centers.

Vj
ij

j th

Cluster center. c Number of cluster centers.


i
th

- Euclidean distance between

and

j th

cluster center.

m = fuzziness index

Figure-1 K-means clustering is the most widely used and studied method among clustering formulations that are based on minimizing a formal objective function. Modification to Kmeans clustering method that makes it faster and more efficient are proposed. The main argument of proposed methods is on the reduction of intensive distance computation, a simple mechanism by which at each iteration, the distance

Figure-2

P a g e | 45

www.ijcait.com

International Journal of Computer Applications & Information Technology Vol. 3, Issue II Aug-September 2013 (ISSN: 2278-7720)

With fuzzy C-means, the centroid of a cluster is computed as being the mean of all points. Weighted by their degree of belong to the cluster. The degree of being in a certain cluster (degree of membership) related to the inverse of the distance to the cluster. By iteratively updating the cluster centers and the membership grades for each data point, FCM iteratively moves the cluster centers to the right location within data set.

6. RESULTS AND ANALYSIS

4. IMAGE ENHANCEMENT
Image enhancement is one of the major research field in image processing. Image enhancement refers to processing an image so that result is more suitable for a particular application. It may be sharpening or de-blurring an out of focus image, highlighting edges , improving image contrast or brightening an image and removing of noise. There are so many techniques to remove noise from an image. Smoothing filters are designed to remove particular type of noise recently neural network turn to be a very effective tool to support the image enhancement.

5. PROPOSED WORK

Figure-4 Figure 4 is a MRI image of human body with scissors inside. This scissor is segmented by using both algorithms K-means and fuzzy C-means clustering for comparison purpose. Number of clusters used is 6.The selection of clusters is done randomly and at cluster 6 we got best results. After MATLAB simulation results are shown below in fig-5 (for k means clustering) and fig-6(for fcm clustering).

Figure 5

P a g e | 46

www.ijcait.com

International Journal of Computer Applications & Information Technology Vol. 3, Issue II Aug-September 2013 (ISSN: 2278-7720)

Figure 6

Thus in proposed method K-means and fuzzy C-means are compared and after simulation we can see that results of both algorithms are same but as already mentioned Fuzzy C-means and K-means are comparative in terms of time. The elapsed time for K-means is 0.302351 seconds and for fuzzy C-means it is 4.40076 seconds. The execution time is less for K means due to less no. of iterations. Thats why K-means is more efficient than Fuzzy C-means.

7. REFERENCES
[1] DIGITAL IMAGE PROCESSING, Rafael C,Gonzalez, Richard E Woods, ADDISION-WELSLEY, An imprint of Pearson Education,3rd edition. [2] Ming-Ni Wu, Chia-Chen Lin and Chin-Chen Chang "Brain Tumour Detection Using Colour-Based K-Means Clustering Segmentation", International Conference on Computer Graphics , 2009. [3] S. Murugavalli and V. Rajamani" An improved Implementation of Brain tumour Detection Using Segmentation Based on Neuro Fuzzy Technique", Journal of Computer Science 3 ,2007 October. [4] S.R. Kannan , Jan. 2005 "Segmentation of MRl using new Unsupervised Fuzzy C means algorithm" ICGSTGVIP Journal, Volume 5, Issue2. [5] Juraj Horvath, 2006 "Image Segmentation using Fuzzy C-means" SAMI 2006. [6] Lei jiang, Wenhui Yang , "A modified Fuzzy C-means algorithm for Magnetic Resonance Images" Proc. VlIth Digital Image Computing: Techniques and Applications, 2003. [7] Fuzzy Logic with Engineering Applications, International edition 1997,-Timothy J. Ross , McGRA W-HILL International Edition.

[8] Y. Zhang, "A survey on evaluation methods for image segmentation, Pattern Recognition, vol. 29, pp. 13 35 1346, 1996. [9] M. Parisa Beham and A. B. Gurulakshmi,Morphological Image Processing Approach on Detection of Tumor and Cancer Cells [10] Sapna Varshney,Navin Rajpal and Ravindar Purwar,Comparative Study of Image Segmentation Techniques and Object Matching Using Segmentation, International Conference on Methods and Models in Computer Science,2009 [11] Pankaj Agrawal,S. K. Shrivastava and S.S. Limaye,MATLAB Implementation of Image Segmentation Algorithms. [12] Ming Luo, Yu Fei Ma and Hong Jiang Zhang. A Spatial Constrained K-means Approach to Image Segmentation. In Proceedings of 4th Pacific Rim Conference on Multimedia. IEEE, 738-742. December 2003. [13] A Tutorial on Clustering Algorithms. Politecnico Di Milano University Press. [Online]. Available: http://www.polimi.it. [Accessed: 2007]. [14] Tranos Zuva, et aI., "Image Segmentation, Available Techniques, Developments and Open Issues," Canadian Journal on image Processing and Computer Vision, vol.2, No.3, March 2011 [15] Rahul Malik, "Combine two clustering technique to improve result," International Journal of Computer Applications & Information Technology, vol.1, No.3, 2012 [16] Deepika N., "Content based image retrieval system using k-means clustering technique," International Journal of Computer Applications & Information Technology, vol.3, No.1, March 2013

P a g e | 47

Anda mungkin juga menyukai