图像修补的优化变换方法  

Optimization Transformation Method for Image Inpainting

作　　者：聂笃宪[1] 李杰[2] 陈鹤峰[3] 

机构地区：[1]华南农业大学理学院,广州510642 [2]湘潭大学土木工程与力学学院,湖南湘潭411105 [3]广东工业大学数学学院,广州510090

出　　处：《计算机工程》2011年第16期232-234,共3页Computer Engineering

基　　金：国家自然科学基金资助项目(50775079);国家青年基金资助项目(60702030);华南农业大学校长基金资助项目(2009K023)

摘　　要：采用整体变分(TV)模型修补图像,提出一种图像修补的优化变换方法。引入一个辅助变量,利用优化变换,将TV模型中单变量函数的优化问题转化为等效双变量函数的优化问题,并利用交替迭代最小化算法和Chambolle’s投影算法求解模型。实验结果表明,与采用梯度下降法的TV模型算法相比,该方法的图像修补效率和修补效果较优。This paper applies Total Variational（TV） model to inpaint image,proposes an optimization transformation method for image inpainting.It uses an optimization transformation technique which involves replacing their univariate functional of TV model by a bivariate functional by adding an auxiliary variable.The Chambolle＇s projection algorithm and alternating minimization method are employed in solving the model.Experimental results show that compared with the algorithm of solving TV model based on gradient descent method,the efficiency and quality of inpainted images by the proposed method is improved.

关 键 词：整体变分 图像修补 优化变换 交替迭代最小化算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]