A Research Of Computational Imaging Algorithm Based On Optical Computing

Computational imaging technology can break through the limitatioins of imaging distance,imaging volume and power consumption of traditional optical imaging systems,and has broad application prospects and development potential.However,with the increase of the scale of image matrix,the problem of insufficient computing power seriously restricts the popularization of computational imaging technology in practical application.Optical computing is an effective way to accelerate matrix computing.With the characteristics of reconfigurability,flexible dimension and direct processing of optical signals in space,the optical simulation computing scheme based on spatial optics perfectly meets the needs of computational imaging algorithm.Therefore,this paper optimizes the computational imaging algorithm based on the spatial optical computing platform,and provides a new idea for realizing high-speed and large-scale computational imaging algorithm.The main work is as follows:Firstly,a spatial optical computing platform based on complex media is built.The process of matrix operation is simulated through the open data set,and the factors affecting the accuracy of matrix calculation are analyzed according to the simulation results.Moreover,the algorithm for calculating the transmission matrix of complex media is studied.Secondly,a random singular value decomposition scheme based on optical computing is proposed.Singular value decomposition is a common means of computational imaging,but it is difficult to decompose large-scale image matrix.In this paper,the inherent properties of complex media are used to reduce the dimension of the matrix.There is no need to generate and store random Gaussian matrix,which can effectively reduce the computational overhead of RSVD.The experiment concludes that at sample raio 0.2 macropixel size 10×10,the relative error of RSVD can be less than 0.1 when 220 grid ground-glass is used as a complex medium.And According to the experimental results,the relationship between quantization accuracy,sampling rate,complex medium and decomposition effect is analyzed.Thirdly,a single pixel imaging algorithm based on optical computing platform is proposed.The measurement process and the data processing process of single pixel imaging are optimized,which alleviates the computational power problem restricting the development of single pixel imaging to a certain extent.The feasibility of the scheme is verified by experiments.And the effects of sampling rate,image pixels,wavelength interval on the accuracy of image reconstruction are analyzed according to the experimental results.