Algorithm Research And Of Image Inpainting Based On Sparse Representation

With the continuous development of digital information technology,the pixel resolution of images is getting higher and higher.However,the image is easily damaged during the process of acquisition,transmission,compression,and storage.Therefore,image restoration is very important.Image restoration technology is also one of the research hotspots in the field of image processing and computer vision.This thesis mainly focuses on researching image restoration based on sparse representation,and on this basis,it improves the image restoration method of sparse representation: Differentiate image block Texture structure information use the structure sparsity,Construct adaptive group structure,Sparse representations and dictionary learning,and finally repair broken images.The main research work is as follows:(1)An image inpainting algorithm based on self-adaptive group sparse representation is proposed to solve the problem that the texture and structure clarity are poor in the repair results.Due to the difference of texture and structure information in the natural images,in order to distinguish the group structure with the original algorithm fixed image block size,firstly,we propose a method to adaptively select the size of the sample image patch to construct an adaptive group structure.Secondly singular value decomposition in groups,obtaining an adaptive learning dictionary of the image patch group,and use the Split Bregman Iteration algorithm to solve the objective cost function.Finally,the adaptive dictionary and the sparse coding coefficient of each group are updated by adjusting the number of image patches and iterations in the group,to get a better restoration effect.The experimental results show that this method not only improves the Peak Signal to Noise Ratio and Feature Similarity Index for Image,but also improves the repair efficiency.(2)For the problem of structural detail information is easily ignored,when extracting image features with the texture information more complex image,resulting in poor repair effect.A new method is proposed: during the group-based learning in the dictionary,use the block structure sparsity algorithm to classify image blocks with prominent texture information and prominent image blocks with structural information in image block groups respectively,divided into two categories.According to the texture information and structure information of the two classes,the dictionary learning is performed respectively,and the overcomplete dictionary with more prominent texture features and more prominent structural features is obtained.And based on this experiment,to ensure that the increase in the time is not much,the image quality after repair is improved.(3)Using Matlab's GUI Tools to implement an image inpainting System.This system can repair two damaged images with random pixel missing and text missing.It can show the contrast effects before and after the repair,it can show the relationship between peak signal-to-noise ratio and iteration number.