An image inpainting method by structural constraints and sample sparse representation

Abstract

This paper discusses a way of image inpainting by structural constraints and patch sparse representation. First, we complete the image edge information by polynomial curve fitting to constrain the structure. Second, we redefine the priority in order to repair the structure firstly and present a narrow-band model based on patch sparse representation to complete structural information. Last, we do sparse representation of the shifting patch in place of the target patch to improve the reliability of sparse representation in texture portion. Simulation results show that the proposed method can make higher image quality and maintain the linear structure well.