一种基于Mumford-Shah模型的红外图像边缘检测方法  

New method for edge detection of infrared images based on Mumford-Shah model

作　　者：方亮[1] 陆佳佳[1] 叶玉堂[1] 杨先明[1] 成志强[1] 

机构地区：[1]电子科技大学光电信息学院,成都610054

出　　处：《强激光与粒子束》2007年第4期566-570,共5页High Power Laser and Particle Beams

基　　金：成都空军装备部资助课题(H04010501w050206)

摘　　要：Chan-Vese模型是一种优秀的简化Mumford-Shah模型。然而Chan-Vese模型是以两个同质区域为基础建立的,这并不符合红外图像的特点,导致直接应用该模型处理红外图像时可能失败。针对这一问题,提出了一种适用于红外图像边缘检测的改进Mumford-Shah模型,并对该模型中目标边缘的保持、停止准则的建立及算法速度的提高作了详细讨论。实验表明,改进Mumford-Shah模型能够克服Chan-Vese模型在对红外图像边缘检测时不能跨越过渡区域的缺点,有效地检测出目标边缘。Chan-Vese（C-V） model is an excellent model of simplified Mumford-Shah models. As Chan-Vese model is based on two homogenous regions, it causes failure in edge detection of infrared images. A level set method for edge detection of infrared image based on improved Mumford-Shah model is proposed. The schemes of preserving object edges, establishing the stopping criterion and speeding up the algorithm are discussed. The experiments show that this model can overcome the failure of C-V model and detect edge of infrared images efficiently.

关 键 词：MUMFORD-SHAH模型 CHAN-VESE模型 红外图像 边缘检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]