On Selecting an Optimal Number of Clusters for Color Image Segmentation

This paper addresses the problem of region-based color image segmentation using a fuzzy clustering algorithm, e.g. a spatial version of fuzzy c-means, in order to partition the image into clusters corresponding to homogeneous regions. We propose to determine the optimal number of clusters, and so the number of regions, by using a new cluster validity index computed on fuzzy partitions. Experimental results and comparison with other existing methods show the validity and the efficiency of the proposed method.

I. Introduction

Region-based image segmentation consists in partitioning an image into non-intersecting regions such that pixels within a region are homogeneous and pixels from adjacent regions are not. Fuzzy clustering algorithms have become popular, e.g. fuzzy c-means (FCM) [1]. Like any unsupervised algorithm, it requires the number of clusters to be set by the user. This makes FCM unsuited for segmenting image databases unless this number is automatically determined. Cluster validity indexes (CVIs) come to this end and once an optimal number of clusters for a dataset is obtained, the corresponding fuzzy partition is selected as the optimal one [2], [3], [4]. Usual FCM and its spatial version [5] we use are presented in section II. In section III, we recall basic definitions on fuzzy aggregation operators [6] and present a CVI to be used in an automatic segmentation process. Experimental results are given in section IV, comparison to other unsupervised segmentation algorithms is provided. We finally conclude and draw some perspectives in section V.