一种用于视频修复的块匹配方法

A BLOCK MATCHING ALGORITHM FOR VIDEO INPAINTING

视频修复最主要的工作之一是对缺失图像块的匹配替代块的搜索。针对传统块匹配算法具有易陷入局部最优的固有缺陷的不足,提出一种新的块匹配搜索算法。该方法结合新的三步搜索方法和简单高效的三步搜索方法的优点,在简单有效的搜索方法确定了四分之一选区后,在该四分之一选区内加上相应的小步长临域搜索点,以提高对于小范围运动的匹配效果。仿真实验结果表明该方法能够更准确、更高效地搜索匹配块,对于视频修复具有非常重要的意义。

Abstract Matching the missing image block pieces instead of searching them is one of the main jobs in video inpainting. In light of the intrinsic shortage of traditional block matching algorithm that it is easy to fall into local optimum, we present a new block matching search algorithm. It combines the advantages of both the new three-step-search method and the simple and efficient three-step-search method, after one quarter of the selected region to be confirmed by the simple and efficient search method, this new algorithm adds the corresponding small-step neighbour search points in this selected region to improve the matching effect for small-scale movement. The result of simulation experiments shows that the algorithm can search the matching block pieces more accurately and efficiently, and has important significance for video inpainting.