Research On Image Super-resolution Reconstruction Technology Based On Sparserepresentation

In recent years,images have been applied in many areas of real life.As an important means to improve image quality,image super-resolution reconstruction technology has attracted extensive attention of researchers.Sparse representation-based super-resolution technology is a hot issue in the field of image processing,but there are also some shortcomings such as too single learning dictionary and vulnerable to noise.In view of the above background,this paper realizes image reconstruction based on sparse representation.The main research contents are as follows:Firstly,some related theories of image super-resolution reconstruction algorithm are expounded.The principles of sparse representation,dictionary learning and updating are introduced,and the superresolution reconstruction based on clustering is compared with other methods.Secondly,on the basis of sparse representation and compressed sensing,an image super-resolution reconstruction method based on improved sparse dictionary is proposed in this paper.In this method,the improved sparse dictionary is introduced into the objective function,and the L1 norm is applied to both the sparse representation coefficient and the dictionary.The experimental results show that this method can improve the image quality and has stable performance in noise environment.Thirdly,aiming at the problems of single dictionary and image degradation,a super-resolution image reconstruction method based on multi-dictionary and sparse noise coding is proposed in this paper.In the dictionary training stage,image blocks are clustered,and different kinds of sub-dictionaries are trained by using classification results to select suitable sub-dictionaries for reconstructing image blocks.When solving sparse representation coefficients,sparse coding noise is introduced,and sparse coding coefficients of original images are estimated by non-local self-similarity of images,and these estimates are used as constraints to remove the influence of noise.The experimental results show that the algorithm is robust to noise.Then,in order to reconstruct the image in the optimal sparse domain,a super-resolution image reconstruction method based on selective sparse coding is proposed in this paper.In the dictionary training stage,the size and phase of image block gradient are used to cluster,and the mapping function is introduced into the target function of dictionary learning to train dictionary for each category.In the reconstruction stage,the high-resolution estimation of image block is obtained by multiplying the sparse coefficient with the corresponding high-resolution dictionary by selecting the appropriate sparse domain for coding.The experimental results show that the method achieves good reconstruction results in the restoration of image edge features and detail information.Finally,the methods in this paper are summarized and compared,and the future research directions are prospected.