Study Of Recovery Algorithm In Compressed Sensing

With the sharpl y d evelopment of the Information technolog y, p eople have an increasing demand for t he information. W hile the traditional sig nal compression and reconstruction follow shannon-Nyquist sampling laws, that is the sample rate must be at least twice as much as t he hi ghest frequency of the si gnal, it c an be rebuild without distortion. This will gives the signal sampling; transmission and storage bring more and more pressure. Compression Sensing theory is on the assumption of the sparsity or the compressibility of the s ignal. It ma kes the si gnal sa mpling a nd c ompression ste p together, avoiding the waste of resources of the traditional sampling compression. Only using a small amount of samples values it can accurately reconstruct the original signal. And it has been widely applied in image compression and processing applications.This paper describes the theory of Comp ression Sensing. Research and summariz e the e xisting r ecovery algorithms, analyze t he pri nciple of t he al gorithms, and compares the performance of the algorithms from many angles. Considering that the traditional r ecovery algorithms ig nore the str uctural information of th e sig nal, so introduces the method b ased on the model, us ing the structur e ch aracteristics of the signal to determine position of the zero elements or larger coefficient in the signal. So that can be used t o more accurately reconstruct the original signal. And com bine it to the a lgorithms of the SP a nd CoSa MP, analyze and com pare t he pri nciple oft he algorithms of the Mb-SP and Mb-CoSaMP.In or der to impr ove the sig nal r econstruction e ffect, r ecovery a lgorithm of the traditional Compression Sensing theory uses wavelet transform to increase signal sparse degrees on spars e tran sform part. But w avelet tr ansform h as the p oor dir ection selectivity. Considering the translation i nvariance of the double tree co mplex wavelet transform (DT-CWT), the latter part of the paper introduc e t he structural characteristics and outst anding advantages of th e DT-CW T. And combi ne it to the algorithms o f the Mb-SP、RAMP and Mb -CoSaMP, put forw ard to the improved recovery algorithm based on the DT-CW T. Th e e xperimental r esults s how tha t the improved a lgorithms c an mor e a ccurately recovery th e or iginal sig nal, a nd reconstructed image effects can be improved, computation time reduces.