Abstract:
Due to the sensitiveness of exemplar-based image repairing algorithm to the fill-front, the current popular algorithm is based on the isophote to decide the priority of reparing. However, the latter cannot distinguish well between the image structure and texture such that the repairing result depends on the image mask and work poorly at the regions composed of complicated texture and structure. This paper proposes an improved colors-analysis-based algorithm, in which the priority is determined according to the color variance between the unknown patch and its neighboring patches. The experiment results show that this proposed method reduces greatly dependence on the mask and ensures the better performance of high quality.