Compressive Imaging Based On Approximate Message Passing

Compressed sensing is a new type of information processing method.It takes full advantage of the sparse characteristics of the signal,and randomly samples the discrete samples at a rate far below the Nyquist sampling rate,and then recovers the original signal through a reconstruction algorithm.Compressive sensing mainly includes three aspects: signal sparse representation,linear observation and reconstruction algorithm.The reconstruction algorithm is the core content of compressed sensing.The approximate message passing algorithm is an efficient reconstruction algorithm.This algorithm not only has low computational complexity but also has high reconstruction accuracy.In this paper,the following aspects are studied based on the approximate message passing algorithm:Firstly,the threshold control parameters in the approximate message passing algorithm are analyzed.For the problem that the control parameters are always invariant during the iteration process,an adaptive parameter-based approximate message passing algorithm is designed.The control parameters are processed before and after the threshold function.The self-adaptive change in the relationship not only speeds up the convergence speed but also improves the reconstruction quality to some extent.Secondly,according to the characteristics of approximate message passing algorithm and high-efficiency denoising algorithm which can improve the quality of reconstruction,an approximate message passing algorithm based on clustering dictionary learning denoising is designed.According to the characteristics of image sub-blocks,clustering dictionary learning is studied.The dictionary is more self-adaptive so as to achieve better denoising effect and further improve the reconstruction quality of the overall algorithm.Finally,for the blockiness problem of approximate message passing algorithm used in block compression imaging,a block-based approximate message passing algorithm based on adaptive sampling is designed.The algorithm adaptively allocates the sampling rate according to the measure of the information volume of the image block,balances the difference between the image sub-blocks after reconstruction,reduces the block effect to a certain extent,and improves the reconstruction quality.