Fast and Accurate Image Inpainting for Advanced Video Coders

In this paper, a new still image inpainting method is proposed. The method assumes that the natural images have the fractal property. The main assumption in these images is that each small block of an image can be replaced using another block contained in that image using translation, scaling, and zooming properties. In our method, a very fast and flexible segmentation algorithm is used which plays a considerable role in the performance of the overall algorithm. In contrary with the previous methods which have employed a fixed block shape, content dependent block shapes are used in our method. Low calculation time accompany with the high quality result have made this method as a good choice to be used in the error concealment part of advanced video coders. Index Terms — Fast Error Concealment, Still Image Inpainting, Advanced Video Coders, Internet.