一种基于模糊C均值聚类的图像区域分割方法  

Region-Based Color Image Segmentation with Fuzzy C-means Clustering

作　　者：周咏梅[1,2] 徐德智[1] 阳爱民[3] 柳萍[3] 

机构地区：[1]中南大学信息科学与工程学院 [2]湖南工业大学计算机系,湖南株洲412008 [3]湖南工业大学计算机系

出　　处：《广西师范大学学报（自然科学版）》2006年第4期203-206,共4页Journal of Guangxi Normal University:Natural Science Edition

基　　金：湖南省自然科学基金资助项目(05JJ40101)

摘　　要：提出一种基于模糊C均值聚类的彩色图像区域分割方法。该方法首先选用适当的彩色空间对图像中的每个像素抽取颜色、纹理及空间位置等综合特征;然后,利用模糊C均值聚类方法进行聚类,利用提出的确定最佳聚类簇数的方法,确定聚类簇数、中心等参数;根据每个像素的隶属函度,将像素初步划归不同的组,利用连接原理对图像区域进一步分割,并提供了图像描述特征。实验结果表明,该方法分割效果很好。A method of region-based color image segmentation with fuzzy C-means clustering ts introduced. First, extracts color, texture, and location features from each pixel by selecting suitable color space. Then ,cluster the integrated feature vectors with the fuzzy C-means clustering and decide the number and the center of clusters with the proposed method of selecting optimal clustering number. Each pixel is grouped and labeled according to its membership degree. Finally ,the regions with the same label are segmented again according to the neighbor connection theory for pixels and some features which describe the regions are provided. Experiment shows this method can be applied widely.

关 键 词：模糊C均值聚类 区域描述 图像分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]