吉林大学学报(理学版)

• 计算机科学 • 上一篇    下一篇

基于快速全局模糊C均值聚类算法的脑瘤图像分割

周文刚1, 付芬2   

  1. 1. 周口师范学院 计算机科学与技术学院, 河南 周口 466001;2. 重庆邮电大学 计算机科学与技术学院, 重庆 400065
  • 收稿日期:2014-12-11 出版日期:2015-05-26 发布日期:2015-05-21
  • 通讯作者: 周文刚 E-mail:zhouwengang@zknu.edu.cn

Brain Tumor Image Segmentation Based on Rapid Global FCM Algorithm

ZHOU Wengang1, FU Fen2   

  1. 1. College of Computer Science and Technology, Zhoukou Normal University, Zhoukou 466001,Henan Province, China; 2. College of Computer Science and Technology, Chongqing University of Posts and Telecommunications, Chongqing 400065, China
  • Received:2014-12-11 Online:2015-05-26 Published:2015-05-21
  • Contact: ZHOU Wengang E-mail:zhouwengang@zknu.edu.cn

摘要:

针对经典模糊C均值聚类算法对初始聚类中心过于敏感的缺陷, 提出一种快速全局模糊C均值聚类算法. 该算法采用分阶段动态递增的方式选取初始聚类中心, 避免了随机化设置导致的聚类结果稳定性差问题. 实验分析表明, 改进后的模糊C均值聚类算法在脑瘤图像分割中的聚类效果较好, 多个数据集的聚类准确率也表明, 快速全局模糊C均值算法的聚类稳定性明显提升.

关键词: 脑瘤, 图像分割, 模糊C均值, 聚类

Abstract:

In view of classical FCM clustering algorithm being too sensitive to the initial cluster centers, a rapid global FCM clustering algorithm was proposed. The algorithm uses dynamic incrementally phased selection of initial cluster centers, avoiding the problem of poor stability of clustering results due to random settings. The experiments show that the clustering result of the improved FCM clustering algorithm is better than that of classical FCM in image segmentation of brain tumors, while the clustering accuracy of multiple data sets also shows that the clustering stability of the rapid global FCM algorithm is enhanced
greatly.

Key words: brain tumor, image segmentation, fuzzy C-mean (FCM), clustering

中图分类号: 

  • TP751.1