Research On Image Completion Algorithms Based On Low Rank And Smooth Prior Information

With the rapid development and wide application of multimedia technology,modern sensor,computer communication and network technology,people often need to store,deal with and analyze image data with larger scale and more complex structure.As a hot research topic in the field of image processing,image completion has attracted wide attention from scholars and industry in recent years.Aiming at the application problem of image inpainting,we have studied deeply from the perspective of matrix and tensor respectively.Unlike the previous repair method based on single low rank prior to image,this paper combines the low rank of the image and the smooth prior,and introduces it to the algorithm of image completion.Compared with previous classical patching algorithm,the proposed algorithm achieves better repair results.The full text structure and content are arranged as follows:The first chapter is an introduction,which describes the background and significance of the research of image completion,and gives the current research status related to image completion.Finally,the main work and the organization structure of the full text are summarized.The second chapter is the related theory and the foundation,and briefly introduces the matrix and tensor basic knowledge related to this article.Finally,a brief introduction to the popular alternating direction multiplier is introduced in this paper.In the third chapter,we study the problem of image inpainting from the perspective of matrix,and propose a new algorithm of joint image matrix,which is low rank and smooth prior.The algorithm achieves perfect image completion by minimizing the Schatten-p norm of image matrix and a class of improved two order total variation.The numerical experiment results show that when the missing rate is bigger,the smaller the pis,the better the effect is.When the missing rate is smaller,the bigger the pis,the better the effect will be.By choosing the appropriate pvalue,the algorithm proposed in this section is better than many algorithms.The fourth chapter studies the problem of image inpainting through tensor perspective,and improves a recently proposed SPC(Smooth PARAFAC tensor Completion)algorithm.Specifically,we will introduce SPC algorithm based on a modified two order total variation constraint to replace the first order total variation constraint in SPC algorithm.Compared with the SPC algorithm,the numerical experiment results show that compared with the SPC algorithm,SSIM and PSNR are better than the original SPC algorithm on the two indexes.The fifth chapter summarizes the work of this paper,and prospects for future potential research.