关键词: 图像聚类, 互信息, 类距离, 对称分布, 粒子群优化

Abstract: When Fuzzy C-Means（FCM） is utilized for image clustering，only the gray level information is considered，ignoring the spatial distribution，so the connection between the original and segmented image is not fully used.This paper constructs a new image clustering segmentation method，which is based on the mutual information between the original and segmented image and cluster distance measure optimum of segmented image.The Particle Swarm Optimization（PSO） based on diversity of symmetrical distribution is adopted as optimization method.Simulations with the medical images by the new method reveal that the results have clear and continuous image boundary，keep complete internal image characteristics.Experimental results show that the image segmentation quality is improved obviously.

Key words: image clustering, Mutual Information（MI）, cluster distance, symmetrical distribution, Particle Swarm Optimization（PSO）