交替迭代的变分修复模型及其分裂Bregman算法  

Variational inpainting model based on alternate iteration and its split Bregman algorithm

作　　者：郝岩[1] 冯象初[1] 许建楼[1,2] 

机构地区：[1]西安电子科技大学理学院,陕西西安710071 [2]河南科技大学数学与统计学院,河南洛阳471003

出　　处：《系统工程与电子技术》2011年第12期2749-2754,共6页Systems Engineering and Electronics

基　　金：国家自然科学基金(60872138)资助课题

摘　　要：为对图像的缺损部分进行有效地修复,提出了一种交替迭代的变分修复模型。通过分析新模型的性质,给出一种高效且快速的迭代算法。新方法首先利用交替极小化方法化原问题为两个去耦的次问题,然后对两个次问题再分别利用分裂Bregman方法进行数值求解。实验结果表明,本文所提出的新算法不但修复效果较好,而且修复速度较快。In order to restore the damaged domain in image effectively, a variational inpainting model based on alternate iteration is proposed. By the properties of the new model, an efficient and fast numerical algorithm is introduced. In the new method, two decoupled subproblems are obtained by using an alternative minimization method, and the two subproblems are solved by using the split Bregman method respectively. Experimental results show the proposed new algorithm can not only get the better inpainting effect, but also improve the inpainting speed.

关 键 词：图像修复 变分泛函 分裂Bregman 交替迭代 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]