A Local Spatial Information and Lp-norm based Fuzzy C-means Clustering for Image Segmentation

Fuzzy local information C-means clustering (FLICM) is robust to image segmentation, but its performance is unsatisfied for image segmentation corrupted by intense noise. This paper challenges image segmentation under intense noise by proposing a novel fuzzy C-means clustering. A new fuzzy factor was proposed in the method, in which local spatial information was enhanced by the neighborhood membership. It is helpful to classify different effects of the neighborhood noisy or non-noisy pixel on the central pixel, thus the proposed method greatly improves intense noise robustness. Furthermore, we take Lp-norm stead of L2-norm in the energy function to improve image segmentation accuracy. Experimental results on synthetic and real-world images show that the proposed method achieves good segmentation performance compared to the traditional FCM and its extended methods, especially for images corrupted by intense noise,.