Research On Image Denoising Algorithm Based On Bilateral Weighting

Image noise removal has been widely used in various fields,and it is a classic problem in image processing.It plays a pivotal role in one aspect of image processing.Image noise removal technology can be used to recover images that are corrupted by noise.Image noise removal technology achieves an improvement in image quality in the form of software,which reduces the cost while obtaining high quality images.At the same time,for the problem that the realistic noise and the mixed noise are too complicated to be processed,weighted image noise removal algorithms are increasingly becoming a popular algorithm in research direction.The main contents of this paper include:First,it is a study to the knowledge of image noise removal,including the basic model of image noise removal,and the classification of image noise removal algorithms.Besides the low rank matrix approximation and the non-local self similarity in image noise removal are discussed in detail.Next,on the basis of studying a variety of advanced image noise removal algorithms,the noise characteristics in real noise images and the image noise removal techniques with weighted nuclear norm minimization analyzed in detail.In this paper,a bilateral weighted nuclear norm minimization algorithm for real color image denoising is studied.Specifically,two weight matrices are introduced into the data fidelity term of the weighted nuclear norm minimization framework to adaptively characterize the noise statistics in each image block of each channel,thereby better utilizing the non-local self-similarity prior of natural images.The proposed real color image denoising model can be re-expressed as a linear equality constraint problem,which is solved by the alternate direction method of the multiplier.Each of the alternate update steps has a closed-form solution and can guarantee convergence,so that the noise-removed image obtained is more effective.At last,aiming at the problem of image noise removal with multiple noises,research on bilateral weighted removal of mixed noise algorithms in sparse models.That is,two weighting matrices are introduced in the data item and the regularization term of the sparse model to characterize the mixed noise and image prior statistics.The pixels are detected by the soft pulse of weighted coding,the simultaneous removal of additive white Gaussian noise and impulse noise is realized.The experimental results show that the denoised image obtained by the algorithm in the image mixed noise removal can better restore the original structure of the image,the information is richer,the robustness is stronger,and the details of the image can be better preserved.