Research On Image Reconstruction Algorithm Based On Compression Sensing

Compressed sensing is a new processing method of information collectionfor compressed signal or sparse signal, the theory has broken the traditionalNyquist sampling theorem, provides a method that it can recover the originalsignal from a small amount of sampling value. When compressed sensingsampling the signal, it also compression the signal, it solves to the contradictioncommendably between signal processing theory of traditional and reality, is abold innovation, displays prominent advantages in the field of modern signalprocessing and has a very broad application prospects.In this paper, after analyzing the existing reconstruction algorithm and therelated optimization theory, summarize the advantages and disadvantages of theexisting algorithm, in the premise of keeping the advantages to find new ways toovercome its shortcomings, to obtain better reconstruction algorithm.First introduces the basic principle of compressed sensing and the basiccontent. It can be divided into the sparse representation of a signal, select theobservation matrix and signal reconstruction three steps. The whole process ofcompressed sensing, the reconstruction algorithm is a key part, but also a focus ofcurrent research.Secondly, combined with the sparsity adaptive matching pursuit algorithmand regularized orthogonal matching pursuit algorithm, researched a new methodfor compressed sensing signal reconstruction, which is called block sparsityadaptive regularization matching pursuit algorithm. The algorithm respectivelysummarized sparsity adaptive matching pursuit algorithm and orthogonalmatching pursuit algorithm’s shortcomings, then combine these two algorithms,in keeping with its advantages, a idea based on block reconstruction is putforward, the method put the two-dimensional image is divided into blocks, toreduce the observation matrix scale of signal single processing, in order to reduce the single processing speed in reconstruction processing, thus the overall runningtime decrease.The last researched a block variable step length before and after the trackreconstruction algorithm based on variable step size before and after trackingalgorithm the existing. The algorithm will also block reconstruction idea ofintegration, not only to meet the requirement of real-time image and the image isexact reconstruction. To solve the problem that the signal reconstruction slow incompression sensing reconstruction algorithm, but also makes the sampling datafor storage and transmission easily.