基于快速搜索最佳匹配块的图像修复算法

Novel image inpainting algorithm based on quickly searching optimum matching block

通过分析Criminisi算法的计算复杂度,得出整个算法的计算复杂度主要取决于其搜索最优匹配块的计算复杂度,且通过分析待修复块优先级的作用,得出整个修复质量与待修复块的优先级密切相关,综合提出了一种QSOMB算法以改善Criminisi算法的缺陷。QSOMB算法一方面采用了一种粗略搜索和精细搜索相结合搜索最优匹配块的算法,可大幅度降低算法的计算复杂度从而节约修复时间,另一方面运用了一种新颖的优先级系数计算方法来确定待修复块的优先级,可得到更为确信的修复效果。通过实验分析可知,相较于Criminisi算法,QSOMB算法是一种有效的且可运用于实践的图像修复算法,其不仅可以确保图像修复后的质量,而且其所需的修复时间更短。

In order to solve disadvantages of the Criminisi algorithm,this paper proposed a novel algorithm named QSOMB algorithm,after analyzing the computation complexity was mainly based on the complexity of searching the best matching block and the quality of the repaired image was almost based on the priority levels of the blocks which was waiting for being repaired.QSOMB algorithm reasonably adopted the both the rough searching and the precise searching together to quickly find the optimum matching block in the image for reducing the computation complexity and saving the inpainting time,and effectively employed a novel fashion to determine the priority levels of the blocks which was waiting for being repaired for guaranteeing the quality of the repaired images. The experimental results show that compared with the Criminisi algorithm,the proposed QSOMB algorithm is effective in both guaranteeing the quality of repaired images and saving the much more repairing time.Thus,it＇s an effective image inplanting algorithm that can be well applied in practice.