摘要
为解决模糊C-均值聚类(FCM)算法在医学图像分割中存在计算量大、运行时间过长以及样本集不理想会导致不好的聚类结果的问题,提出了相应的改进算法.利用收敛速度快的K均值聚类法得到的聚类中心作为FCM算法的初始聚类中心,并将样本对于各个聚类的隶属度之和为1这一约束条件,改变为所有样本对各类的隶属度总和等于样本总数.实验表明,该方法用于人脑磁共振图像分割时,运行速度提高了近3倍,分割准确度明显得到提高.
In order to solve the problems in medical image segmentation using fuzzy c-means algorithm such as large amount of computation, long time of running and imperfect clustering results on condition of imperfect samples, an improved algorithm has been proposed. In the improved algorithm, we take cluster centers obtained by the k-means clustering algorithm as the initial cluster centers for FCM. The restricted condition that the sum of each sample' s membership grades to different classes is 1 is changed into another that the sum of all samples membership grades to different classes is the number of total samples in the improved algorithm. The experiment shows that applying the algorithm to the segmentation of brain magnetic resonance images can obviously improve the speed and accuracy of the segmentation
出处
《华东交通大学学报》
2008年第6期51-54,共4页
Journal of East China Jiaotong University
基金
江西省教育厅科技项目资助(GJJ08237)
关键词
模糊C均值
磁共振成像
图像分割
Fuzzy C-means
magnetic resonance imaging(MRI)
image segmentation