基于隶属度光滑约束的模糊C均值聚类算法  

An Improved FCM Algorithm Using Membership Smoothing Constraint

作　　者：李彬[1] 陈武凡[1] 颜刚[1] 

机构地区：[1]南方医科大学医学图像处理重点实验室,广州510515

出　　处：《中国图象图形学报》2007年第4期623-627,共5页Journal of Image and Graphics

基　　金：国家"973"规划项目(2003CB716101);国家自然科学基金重点项目(30130180)

摘　　要：传统的FCM聚类算法未利用图像的空间信息,在分割叠加了噪声的MR图像时分割效果不理想。本文考虑到脑部MR图像真实的灰度值具有分片为常数的特性,按照合理利用图像空间信息的原则,对传统的FCM聚类算法进行了改进,增加了使隶属度趋向于分片光滑的约束项,得到了新的聚类算法。通过对模拟脑部MR图像和临床脑部MR图像的分割实验结果表明,本文提出的新算法比传统的FCM算法等多种图像分割算法有更精确的图像分割能力,并且运算简单、运算速度快、稳健性好。Fuzzy C-Means （FCM） is a popular clustering algorithm and has been widely used in fuzzy segmentation of Magnetic Resonance（MR） images. However, the segmented results using the conventional FCM when dealing with noisy MR images are not satisfying because FCM takes no spatial information of images into account. Generally an ideal MR images is assumed to be a piecewise constant. We present an improved model of conventional FCM algorithm using membership smoothing constraint. The proposed algorithm can reasonably use the spatial information of images and improve the accuracy of segmentation. The segmentation of simulated brain MR images with different noise level and real brain MR image are presented in the experiments. The results of experiments show that the proposed algorithm is more powerful than many other segmentation algorithms.

关 键 词：图像分割 模糊C均值聚类算法 隶属度光滑约束 图像的空间信息 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]