Image Restoration Method Based On Non-local Thought

With the advent of AI and the booming of the mobile Internet,digital multimedia technology has been applied to all walks of life.Recently,People focus more on image quality,because image has become one of the main carriers of digital multimedia.However,in the process of image acquisition,storage,formation,transmission and pre-processing,the quality of image deteriorates because of some negative factors such as physical,environmental,and shooting.Therefore,image processing technology has caused serious attention of academia.As the basis of image processing technology,image restoration has also become a hot topic of research.Now,image restoration mainly involves three areas including image processing,computer imaging and computer vision.Thus,the research of image restoration has great values both on theory and practice.Image restoration means using some priors of degraded image to restore or reconstruct the image that has been degraded,it also can be known as the inverse process of image degradation.In this paper,the degraded image can be reduced to the sum of the linear shift invariant convolution kernel and Gaussian white noise,the established mathematical model can be expressed as 2)= ? * 1)+ 9),where ‘1)',‘2)' denotes the original image,degenerate image,‘?',‘9)' denotes the point spread function(PSF)and additive noise of image restoration.‘*' means Convolution operator.In this paper,we present a new method of combining non-local methods and matrix decomposition theory for image restoration.Our method is consist of three parts,deconvolution,denoising by bidiagonal decomposition and applying SVD decomposition for final denoising.Experimental results on data show that our method is more effective for restoring blurred images.This article is divided into the following sections.First of all,it introduces the current research situations and evaluation methods of image restoration in detail,then describes the whole framework of the paper.Secondly,it suggests the basic knowledge of image restoration,the degraded models of image,common types of noise and the classic methods of image restoration.After that,the paper proposes an image restoration method based on the combination of non-local ideas and matrix decomposition theory.Finally,the experiments prove that the proposed method can effectively recover a clear image.Then,the paper summarizes the method and gives some prospect for the future.