The Research Of Image Reconstruct Algorithm Based On Compressed Sensing

The traditional signal sampling theory based on Nyquist theorem states that the original signal can be restored completely without distortion if the sampling frequency is higher than two times the highest frequency of signal.With a new perspective,the theory of compressed perception gives a theoretical method to solve the problem.Using the prior knowledge that signals can be sparsely expressed,the original signal can be recovered from fewer observations,which not only breaks the requirement of sampling rate but also relieves the burden on the memory.Reconstruction algorithm is the last step and the key issue of compressed sensing theory.A good algorithm is not only superior to other algorithms in reconstruction time,but also superior to other algorithms in reconstruction quality.This dissertation mainly studies the reconstruction of one-dimensional signal and two-dimensional signal.The main work and research results are as follows:1 For the one-dimensional signal,this dissertation takes the neutron pulse signal as the research object and combines the compression sensing algorithm,and than simulate neutron experiment data.The reconstruction experiment of neutron pulse signal was carried out on MATLAB using Matching Pursuit algorithm,Orthogonal Matching Pursuit algorithm and Compressive Sampling Matching Pursuit Algorithm respectively.The reconstructed signal is analyzed by changing the number of measurements and sparsity.And using signal to noise ratio,peak signal to noise ratio,mean square error,root mean square error to explore the reconstruction of neutron pulse signal under the framework of compressive sensing.The neutron pulse signal in this dissertation is derived from the high-speed nuclear signal measurement system.This research has a certain role in the promotion of nuclear signal research both in theory and in application.2.In order to solve the problems of the existing block compressed sensing and smoothed projected landweber reconstruction(BCS-SPL),this dissertation proposes a block adaptive compression algorithm of image under the double-density dual-tree complex wavelet framework,which solves the fuzzy problem of edge and texture smoothing area.The adaptive design is mainly reflected in the adaptive sampling strategy and adaptive reconstruction strategy,in which adaptive sampling mainly through the standard deviation of the image to measure the texture complexity of the image,adaptive reconstruction is mainly reflected in the filtering operation,The threshold value adaptively changes with the texture information in the filter operation,and the superiority of the algorithm is proved by the experimental simulation.In conclusion,the simulation neutron experiment proposed in this dissertation will be beneficial to the study of nuclear signals,and the experiment has certain significance in protecting national security;at the same time,the BACS-DDDT-CWT proposed in this dissertation has high reconstruction precision and low time complexity,which can be used for reference in the reconstruction of 2d images.