Kernel Possibilistic Fuzzy c-Means Clustering with Local Information for Image Segmentation

Abstract

The kernel weighted fuzzy c-means clustering with local information (KWFLICM) algorithm performs robustly to noise in research related to image segmentation using fuzzy c-means (FCM) clustering algorithms, which incorporate image local neighborhood information. However, KWFLICM performs poorly on images contaminated with a high degree of noise. This work presents a kernel possibilistic fuzzy c-means with a local information (KWPFLICM) algorithm to overcome the noise-related deficiencies of KWFLICM. The proposed approach leverages the robustness to noise of the kernel possibilistic fuzzy c-means (KPFCM) algorithm, which is a hybridization of the kernel possibilistic c-means (KPCM) and kernel FCM (KFCM), rather than relying on the kernel FCM algorithm. Experiments performed on the various types of images degraded by different degrees of noises prove that proposed algorithm is effectual and efficient, and more robust to noise.