几何距优化质心结合隶属度约束RFCM的脑MRI图像分割算法

针对现有图像分割算法聚类复杂以及分割精度不够高的问题，提出了基于几何距优化质心和粗糙模糊C-均值（RFCM）相结合的医学图像聚类分割算法。首先建立软集表示的像素集，并计算每个像素与质心之间的距离，然后基于像素和质心之间的最小距离，将像素分组到聚类中。为了将软集应用到粗糙模糊C-均值中，定义了一个模糊软集，进一步将输入图像转换为二值图像，通过计算连通区域的几何距选择适当的质心。最后利用这些新的质心计算更新像素的隶属度值，从而完成模糊聚类划分。在Allen Brain Atlas等三个医学数据库上评估了所提出混合算法的性能，获得的Jaccards系数和分割精度（SA）都优于几种对比算法。实验证明，提出的聚类分割算法具有良好的性能。

RFCM magnetic resonance brain image segmentation algorithm based on geometric distance optimization of centroid and membership constraints

Aiming at the problem that the existing image segmentation algorithm is complex and the segmentation precision is not high enough, this paper proposed a medical image clustering segmentation algorithm based on geometric distance optimization centroid and rough fuzzy C-means(RFCM). First it set up the set of pixels represented by the soft set and calculated the distance between each pixel and the centroid. Then it grouped the pixels into clusters based on the minimum distance between the pixel and the centroid. In order to apply the soft set to the coarse fuzzy C-means, it defined a fuzzy soft set. It further converted the input image into a binary image, and selected an appropriate centroid by calculating the geometric distance of the connected region. Finally, using these new centroids to calculate the membership value of the updated pixel, it completed the fuzzy clustering. The performance of the proposed hybrid algorithm was evaluated on three medical databases, such as Allen Brain Atlas. The obtained Jaccards coefficient and segmentation accuracy(SA) were better than several comparison algorithms. Experiments show that the proposed clustering segmentation algorithm has good performance.