Research On Image Denoising Method Based On Low Rank And Group Sparse Prior Information

Image denoising is one of the most basic research topics in the field of computer vision.With the continuous development of social economy and the continuous progress of science and technology,many industries have higher and higher requirements for image definition and image processing technology.Since the 21st century,the rapid development of the information age requires the improvement of image processing technology in many aspects,such as military wrestling between countries,increasingly complex diseases,medical treatment,aerospace exploration and many other fields.These fields are inseparable from image processing.In order to meet the needs of higher quality images in daily life,medical treatment,military and aerospace,more effective and time-saving methods are needed to process the images.The first task is to analyze the images,and then denoise the natural images.This paper mainly studies the Non-local means denoising algorithm,low rank denoising algorithm,group sparse denoising algorithm and weight nuclear norm minimization denoising algorithm.The basic research methods are:(1)Fusion group sparsity,low rank and Non-local means and other image prior information,establish the image restoration model for texture detail protection,and study the solution of the model based on convex optimization and non-convex optimization algorithms.Most image restoration models are non-convex optimization problems with variable coupling structure,which are difficult to solve directly.Therefore,we find that it is relatively easy to decompose the original problem into several problems by using alternating direction minimization algorithm convex subproblems,it is a very good solution to solve these sub-problems one by one by using relevant methods.(2)The processing techniques based on low rank and sparse representation are proposed,and the related algorithms are summarized.The noise part of the noisy image is almost in the high-frequency information.By using wavelet decomposition to extract the image information into high-frequency and low-frequency,and using different a priori information to process the different information of the image.The noisy image blocks are clustered to improve the denoising effect.Aiming at the introduction of the related principles of low rank matrix denoising algorithm,the denoising algorithm based on the prior information of image block is studied,that is,the high-frequency information and low-frequency information of the matrix are characterized by wavelet transform,and then the denoising effect is obtained by minimizing the mixed norm and nuclear norm respectively.Finally,taking some pictures as an example,the quality of the processed noisy image is evaluated according to the commonly used evaluation indexes.(3)For different parts of the noisy images,using different prior information,this model has some limitations.Although it has good denoising effect,it will produce some artifacts.On this basis,a new idea was proposed:For a noisy block,we deal with the structure and texture of the sparse coding of the block respectively.Using different norms as regular terms between different prior information will effectively reduce artifacts and improve the denoising effect.(4)Testing the performance of the proposed new algorithm from the two dimensions of subjective and objective evaluation criteria.The subjective evaluation criterion depends on the observation of the object,and the objective evaluation criterion verifies the feasibility of the algorithm by the data presented by structural self similarity(SSIM)and peak signal-to-noise ratio(PSNR).Experiments show that the improved model algorithm has good denoising effect.