Study And Design Of Image Restoration Based On Elastic Network Regularized Binuclear Matrix Decomposition

In recent years,image restoration by low-rank matrix has gradually become a hot spot in image processing.By decomposing the image matrix into lossless image matrix(low-rank matrix)and noise matrix(sparse matrix),the lossless image can be obtained more conveniently and the damaged image can be restored.When the typical low-rank matrix restoration method is used to restore the image,the image effect obtained is much better than the traditional filtering method,and can get a clearer image effect.However,when these methods are applied to more complex images,the effect may be poor,and the calculation cost will increase as the number of iterations in singular value decomposition is higher.Therefore,although this method has been improved in some aspects,it still severely limits its application to large-scale problems and cannot solve practical problems well.Although the proposed algorithm has made a great improvement in image restoration based on low-rank matrix restoration,it still has some problems such as poor effect and long time.On the basis of low-rank matrix image restoration,this paper further expands the research on low-rank matrix decomposition image restoration methods.Aiming at the problem of poor effect and long time in low-rank matrix decomposition algorithm for image restoration,the low-rank matrix restoration theory is studied.Based on this theory,the robust principal component analysis and the optimization algorithm alternating direction multiplier method are further studied.On the basis of the theory is further improved the image restoration of low rank matrix decomposition algorithm,proposed the elastic mesh regularization low rank matrix decomposition image restoration algorithm,this algorithm on the basis of the bilinear matrix decomposition to join F squared norm,get a simple and stable model of bilinear factor matrix RPCA EDN regularization model,In this model,two adjustable parameters are added to improve the effect of image restoration to a certain extent.The effectiveness of the proposed algorithm is proved by two different experimental methods in MATLAB 2018b:(1)Random mask experiment is used to restore the image containing random noise.(2)The block mask experiment will restore the missing images of 2,4 and 8 blocks respectively.By comparing the PSNR value after image restoration and the stability of image restoration effect,it is proved that the proposed algorithm is superior to other algorithms.