基于均值漂移和模糊C均值聚类的图像分割算法  

IMAGE SEGMENTATION ALGORITHM BASED ON MEAN SHIFT AND FUZZY C-MEANS CLUSTERING

作　　者：陈志飞[1,2] 时宏伟[1,2] 吕学斌[1,2] 孙旭[1,2] 

机构地区：[1]四川大学计算机学院,四川成都610064 [2]视觉合成图形图像技术国防重点学科实验室,四川成都610064

出　　处：《计算机应用与软件》2013年第11期13-17,共5页Computer Applications and Software

基　　金：国家自然科学基金项目(60832011)

摘　　要：为改进传统的模糊C均值聚类(FCM)算法应用于图像分割时计算代价大、性能依赖于初始聚类个数和聚类中心、分割过程中易陷入局部极值的问题,提出一种基于均值漂移和模糊C均值聚类的图像分割算法。首先,利用优化的均值漂移算法对原始图像进行分割,分割后形成带权的分割图像并且得到聚类数目和聚类中心;然后,以带权分割图像为输入图像同时把聚类数和聚类中心引入加权FCM算法进行分割;最后,对分割结果进行形态学优化和二值化处理以提升分割效果。实验表明,该方法相对于传统的模糊C均值聚类算法有更好的图像分割效果和分割效率,且分割效果与人类视觉感知具有更高的一致性。To improve the problem of traditional fuzzy c-means clustering （FCM） algorithm that when applied to image segmentation, it has big computational cost, its performance depends on the initial clustering number and clustering centre, and it is easy to fall into local ex-tremum in segmentation process, an image segmentation algorithm based on mean shift and fuzzy c-means clustering is proposed. First, the algorithm uses the optimised mean shift algorithm to segment the original image, after the segmentation there forms the image with the right, and the clustering number and clustering centre are obtained as well. Then, the algorithm chooses the image with the right as the input image, and introduces the clustering number and clustering centre into the weighted FCM algorithm for segmentation. Finally, the algorithm applies morphologic optimisation and binarisation to the segmentation result to improve the segmentation effect. Experimental results show that, com-pared with traditional fuzzy c-means clustering method, the proposed algorithm has better segmentation effect and efficiency, and the segmen-tation effect has a higher consistency with human visual perception.

关 键 词：图像分割 模糊C均值聚类 均值漂移 形态学 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]