Image Reconstruction And Optimization Based On Pseudo-inverse Single-pixel Camera

In contrast to modern cameras,which typically use arrays of millions of pixel detectors to capture images,single-pixel camera can image scenes with a single-pixel detector.Single-pixel imaging modulates the imaging object through a series of illuminating patterns,uses a point detector without spatial resolution to record total light intensity,and finally carries out the correlation operation between modulated lighting sequences and light intensity signal detected by point detector to reconstruct target image.Single-pixel imaging has a broad application prospect,but imaging quality and sampling efficiency greatly limit its practical application.Pseudo-inverse ghost imaging uses pseudo-inverse matrix of measurement matrix to reconstruct image information efficiently.In this paper,a reconstruction algorithm using truncated singular value decomposition to obtain approximate pseudo-inverse of measurement matrix is proposed,combined with pseudo-inverse ghost imaging technique,high quality and high efficiency single-pixel imaging experiments were carried out,which realized reconstruction of high quality images with fewer sampling values and shorter time.The main research contents are as follows:(1)Research on single-pixel imaging of binary objects based on truncated singular value decomposition.Taking binary object as research object,and the measurement matrix is composed of pseudothermal speckles,selecting different truncation parameters for image reconstruction experiment,calculating and analyzing the effects of different truncation parameters on peak signal to noise ratio and structure similarity of images.The results show that high quality images can be reconstructed with appropriate truncation coefficient.(2)Research on pseudo-inverse single-pixel color imaging based on truncated singular value decomposition.Two groups of measurement matrices are selected: one group contains pseudo-thermal speckles subject to exponential distribution and the other group contains simulated patterns subject to binomial distribution.By comparing the two imaging results,it is found that restoration effect of the images is not ideal.Then by applying Gaussian filtering,average filtering,circular average filtering and median filtering to optimize random speckles,pseudo-inverse single-pixel color imaging based on truncated singular value decomposition is experimentally investigated again.The experimental results show that the reconstruction effect is better when optimal speckles are combined with truncated singular value decomposition.In addition,the reconstruction algorithm only needs 0.3second to restore images.Compared with compressed sensing algorithms,the reconstruction time is greatly shortened and the reconstruction efficiency is improved.(3)Research on pseudo-inverse single-pixel imaging of two-dimensional random transformation matrix based on Kronecker product.The projection matrix is constructed by Kronecker product principle,and a theoretical model of pseudo-inverse single-pixel imaging for two groups of random matrices is established.The results of our simulations and experiments prove that the combination of the theory and the reconstruction algorithm in this paper can realize efficient reconstruction of any random speckles.All the above three experimental studies verify that the reconstruction algorithm proposed in this paper can help the target object to realize high-efficient reconstruction,which has very important research significance for the extended application of single-pixel imaging.