纹理图像分割的非局部Mumford-Shah-TV模型及其ADMM算法  

The Non-local Mumford-Shah-TV Model for Texture Image Segmentation and Its ADMM Algorithm

作　　者：杨振宇 潘振宽 王国栋 魏伟波 Yang Zhenyu;Pan Zhenkuan;Wang Guodong;Wei Weibo(College of Computer Science and Technology,Qingdao University,Qingdao 266071)

机构地区：[1]青岛大学计算机科学技术学院,青岛266071

出　　处：《计算机辅助设计与图形学学报》2018年第12期2292-2299,共8页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(61772294);"十二五"国家科技支撑计划项目(2014BAG03B05)

摘　　要：为了提高纹理图像分割的准确率,解决纹理图像中纹理图像成分及纹理区域边界难以描述的问题.基于总变差(total variation, TV)规则项可得到纹理图像区域隐藏的图像结构、非局部算子可以描述纹理图像特征的特点,综合TV模型、非局部Mumford-Shah模型,并用二值标记函数划分区域,提出纹理图像分割的非局部Mumford-Shah-TV变分模型;为了提高计算效率,对所提出的模型设计了相应的交替方向乘子算法,将原问题分解为一系列优化子问题求解.数值实验结果表明,该模型计算的纹理图像区域边界较好,并具有较高的准确率.To improve the accuracy of texture image segmentation,and overcome the difficulties of description of texture components and their boundaries for texture image segmentation,we propose a combined non-local Mumford-Shah-TV model under variational framework making use of the properties of TV(total variation)regularizer in image structure detection and non-local operators in texture descriptions.Meanwhile,a binary label function is used to divide different regions in the model.In order to improve computational efficiency,ADMM(alternating direction method of multipliers)algorithm is designed to decompose the original problems into a series of optimization sub problems.Some numerical examples are presented finally to demonstrate that our model is better in texture image segmentation and accuracy.

关 键 词：纹理图像 图像分割 MUMFORD-SHAH模型 总变差模型 交替方向乘子算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]