图像多相分割的变分水平集方法  

Variational Level Set Method for Multiphase Image Segmentation

作　　者：潘振宽[1] 李华[1] 魏伟波[1] 郭振波[1] 

机构地区：[1]青岛大学信息工程学院,山东青岛266071

出　　处：《系统仿真学报》2008年第S1期417-419,431,共4页Journal of System Simulation

基　　金：教育部新世纪优秀人才支持计划

摘　　要：基于通用的区域模型和通用的区域划分特征函数提出了一种图像多相分割的变分水平集模型,其中区域模型基于图像噪声分布的最大后验概率,特征函数的定义基于水平集函数的Heaviside函数及区域编号的二进制表达。实现了用N个水平集函数划分2~N个区域的划分策略。在能量泛函中增加了水平集函数作为符号距离函数的惩罚项,避免了水平集函数重新初始化的过程。相应的演化方程采用半隐式差分格式进行离散,并给出数值算例进行验证。A general multiphase segmentation formulation under the variational level set framework using general region-based model and general multiphase partitioning characteristic functions is presented.The region-based model is based on Maximum a Posterior probabilities,and the characteristic functions are expressed as Heaviside functions of level set functions and the binary numbers of region labels to implement the scheme partitioning 2~N regions using N level set functions. Additionally,the constraints of level set functions as signed distance functions are enforced in the functional to avoid reinitialization. The corresponding semi-implicit schemes are derived and are used to some examples to validate the method suggested in this paper.

关 键 词：多相 图像分割 变分方法 水平集方法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]