一种与图论Normalized Cut方法相结合的双阈值图像分割  

Image segmentation method by combining double thresholds with Normalized Cut theory

作　　者：朱欣欣[1] 张大志[1] 田金文[1] 张钧[1] 

机构地区：[1]华中科技大学多谱信息处理技术国防重点实验室,湖北武汉430074

出　　处：《红外与激光工程》2006年第z4期48-53,共6页Infrared and Laser Engineering

基　　金：武器装备预先研究基金资助(9140A01060606JW0511)

摘　　要：提出了一种与图论方法相结合的双阈值分割方法,先采用双阈值划分图像区域,然后建立无向图,采用 Normalized Cut 方法划分图,完成图像分割。本文的方法综合考虑了像素间的灰度值和空间距离的相似性,并且运算量比传统的图论方法要小很多。实验结果表明,算法可以取得很好的分割效果。A novel method of image segmentation is proposed that combines,double thresholds with graph theory.The method segments the image into three regions by double thresholds segmentation,then sets up a weighted undirected graph and segments the graph by Normalized Cut.The method has combined the gray level and the position of pixels and reduced the computing time of the traditional graph theory segmentation.Experiments results demonstrate good result of segmentation can be obtained by our algorithm.

关 键 词：图论 Normalized CUT 图像分割 双阈值 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]