Research On Image Restoration Algorithm Of Infrared Image Compressive Sensing Imaging System

The infrared imaging system could realize concealed detection in all-weather of the target area,and it has received extensive attention and a wide application in the fields of people’s livelihood and military.In order to improve the resolution of infrared imaging systems,researchers have introduced computational imaging into infrared traditional optical imaging systems.However,the high computational complexity of the new system leads to a long reconstruction time which severely restricts the practical application of the imaging system.In recent years,some scholars have combined compressive sensing theory with infrared coded aperture imaging system.The new system uses an appropriate image reconstruction algorithm to reconstruct the encoded image data,which could use less sample data than the original image to reconstruct the information of the original image.This article focuses on mainstream reconstruction algorithms,and proposes some new fast reconstruction algorithms to compensate the shortcomings in reconstruction time.Then implement two new algorithms on the infrared compressive sensing imaging system and compare their performance.The main work of this thesis are as follows:First,the article studies the current mainstream greedy algorithms,the methods of minimum l0 norm solution and minimum lp norm reconstruction.Aiming at the problem of long reconstruction time for the lp norm solution method,we study the reconstruction principle and iterative method of iterative reweighted least squares(IRLS)algorithm and propose the affine scaling transformation method to optimize the weighting coefficient of the iterative function.Then to build the fast IRLS(FIRLS_A)algorithm by choosing an appropriate sparse transform.The simulation results show that the FIRLS_A algorithm consumes 26.1%less time than the original algorithm in the reconstruction of infrared images.Secondly,in order to solve the reconstruction problems under complex scenarios environment,we propose the BSC-ONSL0 algorithm and study of the algorithm for solving the minimum l0 norm and the principle of constructing a smooth function.In order to optimize the reconstruction speed of the newton Smoothed l0 norm algorithm,we propose the BSC-ONSL0 algorithm which is based on block compressive sensing theory.And then study the effects of different block sizes on the reconstruction quality and time.The simulation results show that the BSC-ONSL0 algorithm can effectively improve the imaging quality of images under complex backgrounds,greatly shorten the running time,and its average consumptiontime is only 17.6%of the original algorithm.Finally,the prototype of the infrared compressive sensing imaging system is used to test the performance of the two improved algorithms proposed in this paper.The experiment consists of two parts,indoor and outdoor,for investigating the reconstruction performance of the algorithms in different environments.The results of experiment shows that both algorithms reduce the time spent on reconstruction.Among them,the FIRLS_A algorithm is more suitable for infrared human image reconstruction,and the BSC-ONSL0 algorithm is more suitable for complex infrared background image and has a shorter reconstruction time.