关键词: 粒子群, 模糊C均值聚类, 图像分割, 邻域信息, 单元熵

Abstract: In improved fuzzy clustering image segmentation method based on Particle Swarm Optimization（PSO_TDFCM）, the clustering centers searched by particle swarm are taken as image segmentation clustering initializations, which overcomes the sensitive to the clustering center initializations for Fuzzy C-Means（FCM） algorithm as well as improves the speed of FCM algorithm greatly. Meanwhile, on the one hand, the new idea taken into account the great correlation between the spatial site information of a pixel and it’s neighboring pixels, consequently, the neighboring penalized function is added in the objective function；on the other hand, it suggests to update the clustering centers at the two-dimension directions, from which the new objective function combines cell entropy. The results of comparative experiments demonstrate that this approach is an effective fuzzy clustering image segmentation algorithm, which can make a marked improvement in the speed of fuzzy clustering as well as insensitive to the initial clustering patters and robust to the noise.

Key words: particle swarm, Fuzzy C-Means clustering（FCM）, image segmentation, neighboring information, cell entropy