Image inpainting is a challenging problem as it needs to fill the informationof the corrupted regions. Most of the existing inpainting algorithms assumethat the positions of the corrupted regions are known. Different from theexisting methods that usually make some assumptions on the corrupted regions,we present an efficient blind image inpainting algorithm to directly restore aclear image from a corrupted input. Our algorithm is motivated by the residuallearning algorithm which aims to learn the missing infor- mation in corruptedregions. However, directly using exist- ing residual learning algorithms inimage restoration does not well solve this problem as little information isavailable in the corrupted regions. To solve this problem, we introduce anencoder and decoder architecture to capture more useful information and developa robust loss function to deal with outliers. Our algorithm can predict themissing information in the corrupted regions, thus facilitating the clear imagerestoration. Both qualitative and quantitative experimental demonstrate thatour algorithm can deal with the corrupted regions of arbitrary shapes andperforms favorably against state-of-the-art methods.
展开▼
