彩色纹理图像分割的非局部Mumford-Shah多通道全变差模型  

The non-local Mumford-Shah-MTV model for color texture image segmentation

作　　者：杨振宇 潘振宽[1] 王国栋[1] YANG Zhenyu;PAN Zhenkuan;WANG Guodong(College of Computer Science & Technology, Qingdao University, Qingdao 266071 Shandong, P. R. China)

机构地区：[1]青岛大学计算机科学与技术学院,山东青岛266071

出　　处：《重庆大学学报（自然科学版）》2019年第1期110-119,共10页Journal of Chongqing University

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

摘　　要：彩色纹理图像分割的困难在于纹理图像成分的描述及彩色图像层与层之间的耦合。为解决该问题,基于多通道全变差规则项可优化彩色图像层与层之间的耦合,非局部算子可以描述纹理图像特征的特点,提出了彩色纹理图像分割的非局部Mumford-Shah多通道全变差变分模型。所提模型综合多通道全变差模型、非局部Mumford-Shah模型优点,并用二值标记函数划分区域。为了提高数值计算效率,对所提出模型设计了ADMM(alternating direction method of multipliers)优化算法。最后,通过数值实验对比以及定性与定量分析表明方法对于彩色纹理图像的分割取得较好结果。In order to overcome the difficulties of description of texture components and coupling between layers of color image for color texture image segmentation,we proposed a combined non-local Mumford-Shah-MTV model. This model is under variational framework making use of the properties of MTV(Multi-channel Total Variation) regularizer in image coupling between layers 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, we designed the ADMM(alternating direction method of multipliers) algorithm for the proposed model. The results of numerical experiments and qualitative and quantitative analysis demonstrate that the non-local Mumford-Shah-MTV model can obtain better characteristics for color texture image segmentation.

关 键 词：纹理 图像分割 非局部 多通道全变差 ADMM算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]