Comparative Study of Clustering Based Colour Image Segmentation Techniques

Image segmentation is very essential and critical to image processing and pattern recognition. Various clustering based segmentation methods have been proposed. However, it is very difficult to choose the method best suited to the type of data. Therefore, the objective of this research was to compare the effectiveness of three clustering methods involving RGB, HSV and CIE L*a*b* color spaces and a variety of real color images. The methods were: K-means clustering algorithm, Partitioning Around Medoids method (PAM) and Kohonen's Self-Organizing Maps method (SOM). To evaluate these three techniques, the connectivity(C), the Dunn index (D) and the silhouette width (S) cluster validation techniques were used. For C, a lower value indicates a better technique and for D and S, a higher value indicates a better technique. Clustering algorithms were evaluated on natural images and their performance is compared. Results demonstrate that K-means and SOM were considered to be the most suitable techniques for image segmentation among CIE L*a*b* and HSV colour spaces.