A Nonlocal Denoising Framework Based On Tensor Robust Principal Component Analysis With L_p Norm

With the popularization and rapid development of computer network technology and digital multimedia technology,image denoising has gradually become a very important basic subject in computer vision and image processing.Low-rank tensor restoration is widely applied to image denoising.Because the original low-rank tensor restoration model involves l0 norm and rank function,most of the models are NP-hard.The most common solution is to approximate the original low-rank problem to a convex optimization problem.Theoretically,based on the convex optimization method,the real solution of the problem can be accurately recovered only if some conditions are met.But in real life,these strict conditions are often difficult to meet.This means that the method based on convex optimization may not be able to get accurate low rank sparse solutions.In order to solve these problems,this paper summarizes the main contributions and innovations of this paper.(1)A nonconvex low-rank operator recovery model based on l_p norm(l_p-TRPCA)is proposed.The model uses the low-rank terms of norm constraint tensor and the sparse terms of singular value vector to balance the validity and solvability of the model.The optimization algorithm and process of the model are also given.(2)The l_p-TRPCA model is applied to the denoising problem,and a non-local denoising framework for color images and videos is presented.It can deal with zero mean Gaussian noise,impulse noise and any other noise generated by mixing them simultaneously.At the same time,it uses nonlocal denoising strategies to improve the effectiveness of the denoising framework.In addition,we use a nonconvex constraint method.This method can obtain more accurate low-rank tensor recovery results and further enhance the denoising effect of the frame.In the experimental part,we have carried out two sets of experiments to remove Gaussian noise in color images and Gaussian impulse mixed noise in color videos.The experimental results show the effectiveness of the denoising framework.