改进优先级的分步匹配图像修复算法

Double-Step Matching Image Restoration Algorithm with Improved Priority

首先对Criminisi算法的优先级进行了改进,将图像的局部亮度方差作为优先级的一个度量因子,使图像的修复顺序更加合理;然后对Criminisi算法最佳匹配块的获取过程进行了改进,先后使用1范数和最小二乘法,改进了相似性度量函数,进行分步筛选,获取最佳匹配块,使得匹配更为准确,修复效果更加理想。

Firstly, the priority in Criminisi algorithm is improved by adding the local luminance variance of the image as one metric factor of the priority, which gives more reasonable repair order to the image restoration. Secondly, the block matching process of the algorithm given in this paper is changed from one-step to double-step by use of 1 norm and the least square method, respectively, to modify the similarity measuring function. This double-step block matching process obtains more accurate matching block and better restoration result.