基于结构张量的图像修复方法

摘要

Because the traditional anisotropic diffusion equation for image restoration only considers the gradient size, and produces false edges in color image inpainting, this paper proposed an image inpainting method based on structure tensor. Structure tensor was used as diffusion coefficient which can implement different diffusion processes in different regions. The experimental results show that the new method, in comparison with Total Variation (TV) and BSCB methods, improves the image inpainting results, and effectively inpaints color images.%针对传统各向异性扩散方程在修复图像时仅考虑梯度模的大小,且在修复彩色图像时易产生虚假边缘等缺陷,提出基于结构张量的图像修复方法.将结构张量作为各向异性扩散方程的扩散系数,实现在不同区域有不同的扩散方式.实验结果显示:该方法与整体变分(TV)和BSCB方法相比,提高了图像修复效果,有效地完成对于彩色图像的修复.