摘要
针对传统模糊C-均值(FCM)聚类算法在分割低信噪比图像时准确性较差的问题,提出一种用于MR图像分割的改进算法MS-FCM。针对脑部MR图像相邻像素属于同一分类的模糊隶属度相近的特性,在迭代过程中对隶属度数据集进行滤波,以降低噪声对聚类精度的影响。模拟脑部MR图像和临床脑部MR图像的分割实验证明,该算法可以提高图像分割精度。
Aiming at the problem that the accuracy is low when Fuzzy C-Means(FCM) clustering algorithm segments MR images with low signal-to-noise ratio,this paper proposes a modified algorithm named MS-FCM,and applies it in MR image segmentation.Considering the property that the membership values corresponding to the neighboring pixels which belong to the same cluster are similar,it filters membership data sets in the iterate process to decrease the influence on clustering accuracy caused by noise.Experiments on simulation brain MR images with different level noises and real brain MR image show that MS-FCM can improve the accuracy of segmentation.
出处
《计算机工程》
CAS
CSCD
北大核心
2010年第16期198-199,202,共3页
Computer Engineering
基金
国家"973"计划基金资助项目(30730036)
关键词
图像分割
模糊C-均值聚类算法
MR图像
模糊隶属度
image segmentation
Fuzzy C-Means(FCM) clustering algorithm
MR image
fuzzy membership