A kernelized-bias-corrected fuzzy C-means approach with moment domain filtering for segmenting brain magnetic resonance images

Brain magnetic resonance images (MRIs) have complex intrinsic structures with abundant edges, corners, and fine details. Besides these structural complexities, noise and intensity inhomogeneity in the MR images affect the segmentation accuracy. In this paper, we propose a novel approach called kernelized-bias-corrected fuzzy C-means approach using local Zernike moments (LZMs)-based unbiased nonlocal filtering (KBCFCM-LZM) to improve the segmentation performance of brain MR images. The approach works in two phases. In the first phase, we apply LZM-based unbiased nonlocal means filtering on the MR images that provides better pattern-matching capability under various geometric and photometric distortions in the images to remove the effect of Rician noise. In the second phase, the bias field correction process reduces the impact of intensity inhomogeneity, and the segmentation process is performed simultaneously. Further, the segmentation process in the proposed method is carried out in kernel space, which efficiently deals with the problem due to data outliers and provides faster convergence. Thus, the proposed method provides a framework that effectively yields high segmentation accuracy even under high noise and intensity inhomogeneity levels. A comprehensive comparative performance analysis is provided to demonstrate the superior performance of the proposed approach with the state-of-the-art kernel approaches.