数字破损图像的非线性各向异性扩散修补算法  

Digital Image Inpainting Algorithm for Damaged Images Based on Nonlinear Anisotropic Diffusion

作　　者：张红英[1] 彭启琮[1] 吴亚东[2] 

机构地区：[1]电子科技大学通信与信息工程学院,成都610054 [2]电子科技大学计算机科学与工程学院,成都610054

出　　处：《计算机辅助设计与图形学学报》2006年第10期1541-1546,共6页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(10576027)

摘　　要：首先从局部坐标角度分析整体变分(TV)模型与p-Laplace算子的物理意义,从本质上说明p-Laplace算子的扩散性能优于TV模型,进而提出一种基于p-Laplace算子的图像修补算法.该算法利用p-Laplace算子的非线性各向异性扩散的性能来填充受损区域.与TV修补算法相比,文中算法能快速收敛,并达到更好的修补效果,其综合性能优于TV修补算法.The physical characteristics of total variation （TV） model and p-Laplace operator in local coordinates are analyzed, which shows that the diffusion performance of p-Laplace operator is essentially superior to that of TV model. Then, an image inpainting algorithm based on p-Laplace operator is proposed. This algorithm makes full use of the nonlinear anisotropic diffusion of p-Laplace operator to inpaint damaged images. Experimental results show that the proposed algorithm has better general performance in convergence speed and inpainting quality.

关 键 词：图像修补 整体变分模型 P-LAPLACE算子 非线性各向异性扩散 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]