Image Inpainting Based On Texture Synthesis

Digital image restoration is a technology that relies on the known information to fill the damaged part of the image,and it is difficult to detect the traces of restoration.Digital restoration technology of image has been applied in many fields,such as crack repair of old photos,removal of target objects,restoration of ancient murals and so on.Therefore,it is of great significance to study the digital image restoration technology and improve the quality of image restoration.This topic takes Criminisi algorithm and Patch Match algorithm as the main research objectives,summarizes the research status of these two algorithms at home and abroad,and compares and analyzes the advantages and limitations of the two algorithms.Based on the existing algorithms,this paper mainly completes the following work:(1)Firstly,this paper analyzes the Criminisi algorithm and Patch Match algorithm,studies the parameters involved in the algorithm,summarizes the advantages and disadvantages of the algorithm,and points out the shortcomings of the two algorithms.(2)The priority of the Criminisi algorithm may make the filling order wrong.To solve this problem,edge information is added to the priority function,so that the image will be repaired in the direction of the edge structure first.(3)In order to solve the problem of wrong selection of the most similar block in Patch Match algorithm,SSIM(structural similarity index)is added to the algorithm,and the brightness,contrast ratio and structure of the image block are taken as the new criteria to make the candidate similar block more reliable;In order to make up for the defect that Patch Match algorithm uses fixed sample block size,and solve the problem of poor ability to deal with details,the algorithm adopts the method of automatically selecting sample block size,so that the sample block size can be adjusted adaptively according to the image structure information.This article builds a repair system.Simulation results show that the algorithm proposed in this paper is better than other algorithms in visual effect and objective evaluation index,and has a certain practical value.