基于可变大小模板的改进图像修复算法

Criminisi等人提出的基于模板的图像修复算法是该领域最重要的贡献之一。在保持Criminisi基本算法框架不变的基础上,本文提出了一些改进思想。主要贡献是设计了基于可变大小模板的块匹配程序以使模板搜索更准确灵活。此外,对信度更新方式进行了修正,并结合局部搜索以提高算法的综合性能。实验结果证实了改进的有效性。

An improved inpainting algorithm with variable-size-exemplar

Criminisi et al.proposed an exemplar-based image inpainting algorithm,which constitutes one of the most important contributions in this field.With the fundamental algorithmic framework unchanged,some improving heuristics are proposed in this paper.First of all,a novel flexible variable-size-exemplar-based patch matching procedure is carefully designed.In addition,the confidence updating rule is also modified,which combined with efficient local searching helps enhance the overall inpainting performance.Experimental results confirm the effectiveness of the modifications made.