Fast Super-resolution Reconstruction Of Optical Images Based On Learning-based Compressed Sensing Satellites

The optical satellite image refers to the image data obtained by photographing the ground through cameras,multi-spectral scanners,etc.,during the operation of the artificial satellite.However,it cannot be influenced by external factors during the shooting process,which can make the image become Ambiguity,resulting in Super Resolution Reconstruction(SRR).The technology can be applied to medical images,video surveillance,military remote sensing and other fields.This article focuses on the following aspects of super-resolution reconstruction research:(1)Dozens of sets of high-resolution color images of different sizes are collected.Before the dictionary is trained,it is necessary to pre-process the selected dozens of groups of images.which provides a data source for subsequent dictionary training and image reconstruction.(2)The classical Yang algorithm is used to perform image reconstruction by dictionary training K-SVD and dictionary learning OMP.Because K-SVD and OMP will produce more noise in the image reconstruction process,and the number of iterations is more The optimal solution can only be found.Therefore,this paper will further optimize the K-SVD and OMP algorithms so that the optimal solution converges quickly.(3)Since the effect of the edge reconstruction is not obvious after the optimization of the algorithm,the local edge variance gradient edge estimation algorithm is used to further reconstruct the edge.The results show that the optimized algorithm has a certain improvement in peak signal-to-noise ratio and structural similarity,and the reconstruction time is shorter and the reconstruction effect is better.