A Robust Kernel-Based Fuzzy C-Means Algorithm by Incorporating Suppressed and Magnified Membership for MRI Image Segmentation

Abstract

Bias-corrected fuzzy c-means (BCFCM) algorithm with spatial information has been proven effective for image segmentation. It still lacks enough robustness to noise and outliers. Some kernel versions of FCM with spatial constraints, such as KFCM_S1, KFCM_S2 and GKFCM, were proposed to solve those drawbacks of BCFCM. However, the computational performances of these algorithms are still not good enough, especially for large data sets. In this paper, we adopt suppressed and magnified membership idea to speed the computation performance and propose a robust kernel-based fuzzy c-means algorithm (RKFCM). MRI image experiments illustrate that the proposed RKFCM is better than other algorithms in accuracy and computational efficiency. The RKFCM can exhibit the robustness to outlier, noise and weighting exponent m. Experimental results and comparisons indicate that the proposed RKFCM is a fast and robust clustering algorithm and suitable for MRI segmentation.