基于快速步进法的改进图像修复算法

An Improved Image Inpainting Algorithm Based on Fast Marching Method

基于水平集应用的快进修复算法可以简单、快速且有效地修复数字图像中的破损区域,但该方法的线性和局部特征导致其对边缘的保持能力不够,针对这一问题提出了改进方案。在快速步进法的边界行进中引入各向异性扩散,能更准确地保持等照度线的方向,避免彼此交叉。实验结果表明,改进后的算法能更好地保持等照度线的平滑及尖锐边缘,同时具有原算法运算速度快的优点。

The fast marching inpainting algorithm based on level set application can inpaint the damaged regions simply, quickly and effectively, but weak to preserve the edge information because of its linear and local character. An improved algorithm for the above question is presented. Inserting the anisotropie diffusion in fast march method （FMM） boundary evolution is better able to preserve the isophotes＇ directions and prevent the isophotes from crossing each other. The results of experiments shown that the improved algorithm can not only preserve isophotes＇ s smoothness and sharp edge better, but also possess the character of fast algorithm as original FMM.