KFCM Algorithm with Weighted Membership for Brain Tissue Segmentation of MR Image

Medical images are affected by imaging mechanisms and inevitably contain noise.In order to solve the problem that traditional medical image segmentation algorithms are sensitive to noise,this paper proposed an improved KFCM segmentation method based on weighted fuzzy membership degree.In this method,we defined a local membership function which introduces the local spatial information based on the traditional KFCM algorithm.Then,the weighted membership function is constructed by combining the proposed local membership function with the global membership function from the traditional KFCM algorithm to calculate the membership value for each pixel.Finally,the membership value of each pixel is redistributed by iterative aggregation based on local information.The experimental results on Simulated Brain Database with different noise demonstrate that our method can improve the segmentation accuracy while ensuring the robustness to noise.