Enhanced kernel-based fuzzy local information clustering integrating neighborhood membership

To enhance the segmentation performance and robustness of kernel weighted fuzzy local information C-means(KWFLICM) clustering for image segmentation in the presence of high noise, an improved KWFLICM algorithm aggregating neighborhood membership information is proposed. This algorithm firstly constructs a linear weighted membership function by combining the membership degrees of current pixel and its neighborhood pixels. Then it is normalized to meet the constraint that the sum of membership degree of pixel belonging to different classes is 1. In the end, normalized membership is used to update the clustering centers of KWFLICM algorithm. Experimental results show that the proposed adaptive KWFLICM(AKWFLICM) algorithm outperforms existing state of the art fuzzy clustering-related segmentation algorithms for image with high noise.