中心线模式Snake算法对细长目标的提取  

Slender Object Extraction for Image Segmentation Based on Centerline Snake Model

作　　者：刘洪江[1] 汪仁煌[1] 何最红[1] 李学聪[1] 

机构地区：[1]广东工业大学自动化学院,广州510090

出　　处：《光电工程》2011年第9期124-129,共6页Opto-Electronic Engineering

基　　金：国家自然科学基金资助项目(11072063)

摘　　要：针对有强干扰的细长目标提取,本文提出用中心线和宽度曲线演化来提取目标的Snake算法,该算法利用细长目标关于中心线对称的特性,用中心线和宽度曲线两个一维函数代替表示目标边缘的二维函数,通过控制这两个一维函数的形变来使Snake模型能量最小,达到准确提取目的。算法还利用了边缘强度对称性,增加图像力加权函数,能快速避开非对称性边缘的干扰,减少迭代次数。新算法减少控制点数量,降低计算规模,能更快获得结果,能实现全自动分割。实验结果表明该算法具有较强的抗干扰性,能使分割细长目标效果良好。To extract slender object from image with strong interference,a method based on Snake is proposed by using centerline and width＇s curve.Making use of the centerline symmetry of slender object,the contour is described by using two one-dimensional functions about centerline and width instead of two-dimensional expression.By evolving this two functions to meet the minimization energy of snake model,the object is accurately extracted.The symmetry of edge gradient is used,and a weighting function is mentioned to increase the image force.By this way,the interference can be avoided and iterations can be reduced.The improved algorithm reduces the number of control points and the size of calculation,and the results can be obtained automatically and fast.Experiments show that the improved algorithm is effective.

关 键 词：对称 中心线 蛇状模型 图像分割 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TN911.73[自动化与计算机技术—计算机科学与技术]