earticle

영어

Exemplar-based image inpainting method is widely used for recovering the missing regions, especially when the missing regions are large. The method fills in the target region patch by patch, and a target patch is compared to each candidate patch in whole image to get a best one. In this paper we propose a method to speed up the best candidate searching process. Inspired by the fact that the patches near to a target patch are more likely to be the best candidate than the farther ones, the similarity measure is modified to cooperate with the distance weighting and bounded search is adopted. It has been shown by experiments that the proposed method greatly reduces the number of pixel error computations.