Research On Super-resolution Reconstruction Algorithm Based On Prior Model

In the process of acquiring images,due to the defects of imaging equipment process,imaging distance,atmospheric interference,motion and other factors,the image is degraded and the use value of the image is affected.Based on the existing imaging equipment,Super-resolution technology can improve the spatial resolution of image through software post-processing methods,which has been favored by researchers.Super-resolution technology uses low-resolution images to estimate highresolution images,which can recover the lost high-frequency information,and improves the spatial resolution of the images.That technology has been widely used in various fields of image processing,such as medical images for disease diagnosis,video surveillance for surveillance security,and satellite images.This paper analyzes the background and research significance of super-resolution technology,systematically summarizes and elaborates the main realization algorithms of super-resolution technology,and introduces commonly used image quality evaluation methods.Aiming at the ill-posed nature of the super-resolution reconstruction problem,the reconstruction process needs to add priori knowledge to constrain the solution and transform the ill-posed problem into a well-posed problem.Therefore,this paper focuses on the influence of different prior models as constraints on the reconstructed image.The main research work is as follows:(1)Aiming at the poor performance of the denoising performance of the prior model constrained by the image gradient L1 norm in flat areas,this paper proposes a joint L1 and L0 prior model.In this paper,the model constrains both the edge region and the flat region,which preserves both the edge-preservation of the L1 prior model and the denoising of the L0 prior model in the flat region.In the reconstruction process,the joint estimation method is selected to register the image,that is,the high-resolution image is updated while the motion deformation matrix registered to this high-resolution image is updated to reduce the error caused by the registration accuracy;under the framework of maximum a posterior probability method,the mixed prior model is used to constrain the solution to ensure its uniqueness and stability;the MajorizationMinimization(MM)algorithm is selected to solve the constraint term of the L1 norm,and the half-quadratic splitting L0 minimization method is selected to solve the constraint term of the L0 norm,and the iterative process of the half-quadratic splitting L0 minimization method is improved according to the image gradient histogram distribution to make the image sparser.Experiments were conducted on simulated image sequences.From the comparison of image quality evaluation and visual effects,it can be seen that the algorithm in this paper effectively suppresses noise in flat areas while preserving edges.(2)Aiming at the problem that the algorithm in this paper cannot realize the adaptive estimation of all parameters,this paper starts from the definition of the L0 norm,analyzes the relationship between the intermediate variables introduced in the half-quadratic splitting L0 minimization method and the L0 norm parameters,and gives the adaptive solution formula of this parameter to realize the adaptive estimation of all parameters.The experimental results of the real image sequences show that the reconstruction algorithm based on adaptive parameter estimation can reconstruct a clearer image.