Research On Super-resolution Reconstruction Algorithm Of Remote Sensing Images Under The Framework Of Compressed Sensing

With the development of image processing technology,people’s requirements for image quality are getting higher and higher,and detailed and clear high-resolution images are of great significance in many fields.However,restricted by the current manufacturing process and hardware cost,the acquired images often cannot meet the actual application requirements.How to improve the image resolution without increasing the hardware cost has become a hot research topic,and the use of image super resolution reconstruction technique is one of the solutions.This technique is a typical pathological inverse problem in mathematics,which needs to add some prior knowledge about the imaging process or image model as a constraint to the solution space using a regularization method when solving.Compressed Sensing theory,which exploits the sparse prior of the signal and can reconstruct the original signal accurately at much smaller sampling counts than those set by Nyquist’s sampling theorem,shows strong constraining power as a prior knowledge in the image inverse problem.In this paper,based on the theory of compressed sensing,the problem of super-resolution reconstruction of remote sensing images is studied on the basis of using image local sparsity a priori,and two effective super-resolution reconstruction algorithms are proposed,and the main research results are as follows.1.To obtain better image super-resolution reconstruction results,an image super-resolution reconstruction algorithm based on adaptive weighted compressed sensing,called the AWGPSR algorithm,is proposed.The algorithm considers the local sparsity of the image,assigns different weights to each element in the regularization term in the image SR optimization model,and changes them adaptively in each iteration according to the results;in addition,the GPSR algorithm is used instead of the traditional OMP algorithm in sparse coding to overcome its drawback of requiring a fixed sparsity.Through the above two means,more accurate sparse representation coefficients can be obtained.Meanwhile,considering the complex texture characteristics of remote sensing images,only the high-frequency components of the images are reconstructed in super-resolution to ensure that the image SR is concentrated in the edge and texture-rich regions.The experimental results show that the algorithm can effectively perform super-resolution reconstruction of remote sensing images with clear texture details and small reconstruction errors.2.A compression-aware adaptive regularization term-based image super resolution reconstruction algorithm,called the ARSR algorithm,is proposed.The reconstruction effect of an image is closely related to the sparse representation coefficients.To further obtain more accurate sparse representation coefficients,the algorithm adds a regularization term after the optimization model using the above adaptive weighting method to take the image correlation into account,and this regularization term can adaptively reconcile the relationship between sparsity and correlation by generating a suitable coefficient.In addition,an ADMM method is used to derive an approximation to solve the proposed optimization model effectively.The experimental results show that the algorithm reconstructs better than several existing classical image super-resolution reconstruction algorithms.