Research On Image Inpainting Algorithm Based On Texture Information For Object Occlusion

Image inpainting is an image processing technology based on information from known parts of the image to restore the information part of the missing part.It has been widely applied to many aspects,such as cultural relic recovery,text removal,movie and TV special effects.The study of image inpainting technology is of practical application value.This paper first provides an overview of image inpainting research background and significance and research status at home and abroad,introduces the image inpainting algorithm based on classical models,including Total Variation(TV),Curvature Driven Diffusion(CDD)image inpainting algorithm of image inpainting algorithm,Criminisi image inpainting algorithm and Sparse Representation(SR)image inpainting algorithm.Secondly,removal of large area objects,a Criminisi image inpainting method based on wolves algorithm is proposed.The method of priority calculation is optimized by introducing boundary term and regularization function,the optimization module matching algorithm according to the exploration of wolves,the wolf walk,fierce wolf running,fierce wolf siege,wolves update steps,combined with the global search and local search to find the matching of image blocks,and then the image block matching to fill the area to be repaired.The experimental results show that the method has better visual effect,and improves the peak signal to noise ratio(PSNR)and algorithm speed.Then,removal of object occlusion in non-block areas,a sparse representation image inpainting method based on distance clustering structure group is proposed.This method introduces Euclidean distance to characterize the structural similarity between image blocks,and sets blocks of image blocks by setting threshold,and blocks the image blocks with structural similarity.Then image block of packets for sparse decomposition,dictionary learning,optimization,reconstruction to repair the image.The algorithm proposed in this paper is compared with the TV algorithm,the CDD algorithm and the group sparse representation algorithm(Group-based Sparse Representation,GSR).The experimental results show that the method has achieved better image restoration effect,the peak signal to noise ratio is improved and the speed of the algorithm is faster.Finally,the work of this paper is summarized and the development of image inpainting is prospected.