A New Multiple-Kernel Fuzzy C-means Algorithm with Texture and Special Information for Image Segmentation

One of the important issues expressed in image processing is identification and separation of image into its components (segmentation) which determines the final success or failure of image analysis methods. Clustering-based methods are one of the methods used in image segmentation. In these methods, data is grouped such that those included in the same cluster have the most similarity with each other. In this paper, two multi-kernel FCM algorithms with texture and spatial information (MKFCM-S1, MKFCM-S2) are presented for segmentation. In the proposed algorithm, a combination of spatial information and texture information and has been used, so a kernel function for spatial information and another kernel for texture information is considered. Experiments on artificial and real image show that algorithms run time has reduced considerably and convergence has increased in the proposed algorithms.

Keywords:

clustering , Fuzzy C-Means (FCM) , image segmentation , multiple kernel