关键词: 水平集演化, 模糊聚类, 边缘竞争, 自动分割, 脑部磁共振图像

Abstract: A level set segmentation method based on edge competition for automatic segmentation of brain Magnetic Resonance (MR) images was proposed, which employed the Robust Fuzzy C-Means Based Kernel Function (RFCMK) result as priori knowledge to solve the problems of noise and edge leakage. Firstly, the image was pre-segmented with the RFCMK algorithm. Then, the sub-class images derived from pre-segmentation were processed by threshold operation to get the edge, which was used as the initial contour for the level set evolution. Finally, a competition mechanism combining the gradient information of sub-class image and original image was introduced to the edge indicator. An image set from a simulated brain database and a real brain MR image were tested to validate the accuracy of the proposed method. The range of area-based and edge-based statistical value is [0.91, 0.95] and [0.05, 0.22], respectively. The experimental results show that the proposed method can detect the edge accurately, and reduce the effect of noise.

Key words: level set evolution, fuzzy clustering, edge competition, automatic segmentation, brain Magnetic Resonance (MR) image