Effective exemplar-based image inpainting using low-rank matrix completion

This paper proposes a novel image inpainting algorithm which adopts a new way of determining patch priorities and a new strategy to search the matching patches. More importantly, the proposed algorithm takes full advantage of the similar structure of the matching patches and introduces the low-rank matrix completion effectively to repair the damaged region. The experimental results show that the proposed algorithm can set reasonable patch priority, reduce the probability of mismatching, and maintain the completeness of structure. The repaired images have more natural appearance than the results of some other mainstream image inpainting approaches.