文摘
Since most images were built with regular textures and structures, the exemplar-based inpainting technique has become a brand-new solution for renovating degraded images by searching a match patch. This characteristic has also been named as the local self-similarity. Nevertheless, traditional exemplar-based methods try to find the best match patch in the whole image with only one direction; thus, often leading to a non-ideal repairing result. In this article, we propose a novel patch matching technique to rebuild the structure and texture of image, in which the surrounding information of the patch is fully concerned. Aside from determining a more precise filling priority by the sparsity of the image structure, we have applied the difference of fractal dimension to enhance the similarity between the source patch and the target patch. Experimental results have demonstrated the superiority of the proposed technique over related works in the renovating accuracy.