摘要
针对FCM进行图像分割时需要人为确定聚类数的问题,提出一种改进的基于FCM的图像分割算法.该算法先对图像进行4叉树结构的子图分解(即原图等分为2×2的4幅子图,子图再等分为2×2的4幅子图),待子图满足一定条件时进行聚类数为2的FCM聚类分割;然后将分割好的区域根据其大小及相邻区域直方图的巴氏距离进行合并,得到最终的分割结果,从而避免了聚类数目的直接确定.实验结果表明:该算法能够获得很好的分割效果;对子图进行聚类分割减少了每次参与聚类的对象数,从而在一定程度上降低了算法的计算量.
The clustering number needs to be determined artificially when image segmentation using FCM is performed. In order to solve this problem,an improved algorithm based on FCM is proposed. In this algorithm,first, a sub-image decomposition of the ori-ginal image is conducted on the basis of quad-tree structure,by which the original image is divided into 2 × 2 sub-images equally and each sub-image is divided into 2 × 2 sub-images equally again until the sub-image satiates certain conditions. Then,the sub-image is segmented by using FCM with the clustering number 2. Moreover,region merging is carried out according to the region area and the Bhattacharyya distance of two adjacent regions’histogram,with the segmentation results being finally obtained without determining the clustering number directly. Experimental results indicate that the proposed algorithm possesses good seg-mentation effectiveness,and that,to some extent,the computation complexity is reduced,because the segmentation using FCM helps reduce the number of samples for each clustering.
出处
《华南理工大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2014年第3期1-7,共7页
Journal of South China University of Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(61372008)
亚热带建筑科学国家重点实验室开放课题(2013KA02)
关键词
图像分割
信息熵
聚类数
FCM
FCM
image segmentation
information entropy
clustering number
作者简介
周晓明（1963-），男，教授，主要从事电磁场与信号处理研究．E—mail：zhouxm@scut．edu．cn
