Research On Image Super-resolution Reconstruction Algorithm Based On Sparse Representation

Image is one of the main ways for human beings to recognize the world.It plays an important role in daily life and work,and is widely used.However,in the imaging process,the resolution of the image is easy to be reduced by the interference of natural factors such as noise,which will affect the subsequent application.Therefore,how to obtain high resolution image has become a hot issue in image processing.Super-resolution reconstruction technology can use software method to transform low-quality fuzzy image into clear high-resolution image,which has important scientific significance and application value in the field of image processing.In the existing super-resolution reconstruction algorithm,the sparse representation based method has a good reconstruction effect.Through sparse decomposition of the image to restore the lost information,but the algorithm still has some defects in the restoration of image detail information.In order to solve the above problems and further improve the quality of image reconstruction,this paper improves the existing sparse reconstruction algorithm from two aspects of feature extraction and global constraints.Compared with the existing sparse reconstruction algorithm,the improved algorithm can maintain the geometric characteristics of the image more effectively and has higher PSNR value.The main improvements are as follows:1.In order to solve the shortcomings of the traditional sparse reconstruction algorithm that the direction of feature extraction operator is single and the information is not comprehensive enough,a sparse reconstruction algorithm based on multi-directional gradient information extraction is proposed.According to the characteristics of the image block,the algorithm extracts the features of the image in the four directions of transverse,longitudinal and diagonal respectively,and uses sparse model to reconstruct.Finally,the high-resolution image is synthesized.Compared with the original gradient operator,the feature information extracted by the multi-directional gradient operator is more delicate and rich,which can restore the image geometry more completely and improve the quality of image reconstruction.2.In order to further improve the image quality,the global constraint part of the traditional coefficient reconstruction algorithm is improved.A global constraint model based on image self-similarity is proposed,which takes a large amount of redundant information contained in the image itself as the regularization constraint,and uses the near end gradient descent method to solve the convex optimization problem of the global constraint to complete the reconstruction of the initial high-resolution image.The experimental results show that the global constraint model based on selfsimilarity can improve the PSNR value by 0.2db,which can effectively deepen the overall contour of the image and improve the reconstruction effect of the image.