Image segmentation using rough-fuzzy K-medoid algorithm

Recently image segmentation based on rough set and fuzzy set have gained increasing attention. In this article, a rough-fuzzy K-medoid algorithm is proposed for color image segmentation. The main objective of this algorithm is to provide an efficient method which uses color information (R, G, B values) along with neighborhood relationships. In this method K-medoid algorithm is modified using reduct formation rule of rough set theory while membership values of the features are obtained using fuzzy sets. This method uses spatial segmentation where an image is divided into different parts with similar properties. Choice of initial cluster centers affects the performance of K-medoid algorithm, even if it is a simple and effective one. In this article, a modified K-medoid algorithm is proposed having two parts- in the first part, the initial cluster centers are optimized by rough set theory and in the second part the optimal cluster centers are used to execute K-medoid algorithm. The proposed scheme does not require any prior information about the number of segments. Results are compared with five different state of the art image segmentation algorithms and are found to be encouraging.