Super-Resolution Image Reconstruction Based On Local Structure Similarity And Sparse Representation

The method based on sparse representation performs well in image super-resolution reconstruction,but the traditional sparse representation method considers the sparsity between image patches independently,which leads to the loss of part of texture structure in the reconstructed image.In this dissertation,a super resolution image reconstruction algorithm based on local structure similarity and sparse representation is proposed.In this algorithm,the constraint model with similar local geometric structure and the₁L norm regularized sparse representation are used to solve the sparse representation of image patches in the low-resolution dictionary,and the image sparse representation with local structural information is used to reconstruct high-resolution image patches.Experimental results show that this algorithm can recover the image texture better than the bicubic interpolation algorithm and the traditional sparse representation method.The contents of this dissertation are organized as follows:The first chapter introduces the status and research significance of super resolution reconstruction technology in the field of image processing,briefly describes several reconstruction methods and their advantages and disadvantages,and points out the future research direction in the field of super resolution reconstruction.The second chapter introduces the basic principle of sparse representation in detail,including the algorithm principle used in the two stages of dictionary training and sparse coefficient optimization.Chapter Three introduces the similarity principle of image local structure,including the definition of image local structure and a variety of similarity measurement methods.In Chapter Four,an image super-resolution reconstruction algorithm based on local structure similarity and sparse representation is proposed,which introduces the local manifold structure of the image interior space into the sparse coding,and experiments are conducted to verify the superiority of the proposed algorithm.In Chapter Five the whole work is summarized and the future work is simply discussed.