• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

J4 ›› 2011, Vol. 33 ›› Issue (4): 102-106.

• 论文 • 上一篇    下一篇

基于直方图相关性约束的快速多阈值FCM图像分割算法

来跃深,马天明,田军委   

  1. (西安工业大学机电工程学院,陕西 西安 710032)
  • 收稿日期:2010-05-20 修回日期:2010-09-28 出版日期:2011-04-25 发布日期:2011-04-25
  • 作者简介:来跃深(1965),男,陕西西安人,教授,研究方向为计算机控制与测量。马天明(1981),男,陕西西安人,硕士生,研究方向为图像分割和图像处理。田军委(1973),男,博士,副教授,研究方向为视觉测量、模式识别与机器视觉、智能电子信息系统、精密与超精密伺服控制与检测等。
  • 基金资助:

    陕西省教育厅专项基金资助项目(09JK497)

Fast MultiThreshold Fuzzy CMeans Image Segmentation Based on Histogram Correlation Constraints

LAI Yueshen,MA Tianming,TIAN Junwei   

  1. (School of Mechatronic Engineering,Xi’an Technological University,Xi’an 710032,China)
  • Received:2010-05-20 Revised:2010-09-28 Online:2011-04-25 Published:2011-04-25

摘要:

针对传统的模糊C均值(FCM)聚类算法在样本数和特征数较多时,运算较为复杂以及耗时较多的问题,本文提出了一种采用直方图的相关性作为约束采样率的快速多阈值FCM分割方法,控制图像失真,使得需要运算的数据量减少,以获得较快的分割速度。由于借助了基于模糊集的图像分割技术——模糊C均值算法实现多阈值图像分割,考虑到了每个像素对于聚类中心的隶属度,使得其有较好的适用性。根据实验结果,在保持传统FCM算法的分割效果的前提下,该算法的分割灰度图像耗时是传统FCM的1.4%,因此该算法具有一定的应用价值。

关键词: 模糊C均值聚类算法, 图像分割, 模糊聚类, 直方图, 相关性

Abstract:

The traditional fuzzy Cmeans (FCM) clustering algorithm has some problems, such as massive calculation and slow operation speed, especially the large amount of data. A fast multithresholds FCM algorithm based on histogram correlation constraints is proposed to control the image distortion due to resampling. Because of the amount of data in the operation has been reduced,the segmentation speed turns faster. In this paper, image segmentation uses the  fuzzy techniques of the fuzzy C Means (FCM) algorithm which considers each pixel for the cluster center membership. FCM can achieve multithreshold image segmentation which features good applicability. The experimental results which make it valuable on application shows that the proposed algorithm preserves the effect and costs only 1.4% the time of the traditional FCM.

Key words: fuzzy Cmean (FCM) clustering algorithm;image segmentation;fuzzy clustering;histogram;correlation