压缩感知重建算法的若干分析-信号与信息处理专业论文.docxVIP

AbstractCompressed Sensing(CS)theory is a new signal sampling theory proposed in recent years for for the sparse signal or compressible signal. It is a bold innovation of signal sampling theory which broke the bottleneck of traditional Nyquist sampling law and have a very broad application prospects. The research of reconstruction algorithms is a very important direction in CS field.In this paper,we focus on the survey of reconstruction algorithms. Specific content and innovations are as follows:(1)A new matching pursuit algorithm named BSARMP (Blocking Sparsity Adaptive Regularized Matching Pursuit) are given.On one hand, this method combined advantages of existing ROMP algorithm and SAMP algorithm.On the other hand, we divide target image into suitable small blocks, the speed to reconstruct single block image reduced, thus reduced the processing speed of the entire image. The experimental results show that as long as the block is reasonable, the algorithm has better time performance and the reconstruction results.(2)A gradient project algorithm is given in this paper named EGPSR(Equalized Gradient Projection of Sparse Reconstruction). And based on this, EBGPSR(Equalized Blocking Gradient Projection of Sparse Reconstruction) algorithm is proposed. EGPSR method improved the reconstruction quality compared with GPSR algorithm by processing the target matrix both by column and row . And EBGPSR algorithm bring the idea of blocking into EGPSR method and improved the speed to reconstruct the image.Simulation experiments show that the EGPSR algorithm has a better reconstruction effect, while EBGPSR algorithm based on EGPSR algorithm can improve the reconstruction speed.(3)A new algorithm named MFRCG(Modified Fletcher-Reeves Conjunction Gradient)is given in this paper. New algorithm is a method with simpler iterative process and less storage space requirement.The experimental results show that MFRCG algorithm can get higher peak signal-to-noise ratio (PSNR)compared with FR