Research On Diffraction Spectrum Computational Imaging Reconstruction Technology

As an important part of computational optical imaging,diffractive spectral computational imaging technology has important application value in the field of optical remote sensing.Unlike traditional imaging spectrometers,diffractive optic image spectrometers exploit diffractive optical elements to achieve both dispersion and imaging,which have the characteristics of large luminous flux,compact structure,high cost performance,gaze imaging,easy miniaturization,and high stability.However,in the process of data acquisition,infocus images are often blurred by other defocused images.How to recover clear images from heavily polluted spectral images has become a bottleneck restricting the development of diffractive spectral computing imaging technology.For such ill-posed inverse problems about removing cross-channel blur,the restoration effect of existing reconstruction algorithms is not ideal.In this paper,we focus on the technology of diffractive spectral computational imaging,and conduct in-depth research on diffractive spectral image reconstruction.The main works of this paper are as follows:(1)In this paper,the imaging mechanism of the diffractive optical element and the reconstruction process of the 3D diffractive spectral images are expounded in detail.In addition,two theoretical models of the point spread function and solutions to keep the lateral magnification of the imaging system constant are discussed.Based on the principle of diffractive spectral imaging,a mathematical simulation model of image degradation process is established,and a narrow-band diffractive spectral imaging system is set up.Moreover,the advantages and disadvantages of the existing reconstruction algorithms are further studied,which lays a solid theoretical foundation for subsequent algorithm improvements.(2)An algorithm for reconstructing diffractive spectral images based on the multichannel spectral-spatial total variation model is proposed.This method comprehensively considers the spatial and spectral prior information to construct a spatial-spectral total variation regularization term,and imposes local spatial smoothness constraints and local spectral smoothness constraints on the reconstruction results.The regularization term can also adaptively adjust the denoising intensity of each band according to the spatial and spectral information.Because of the spectral total variation,the ill-condition of the problem is mathematically reduced,so that the proposed method can obtain a stable approximate solution.In addition,the method proposed in this paper uses the alternating direction method of multipliers and fast Fourier transform algorithm to decompose the proposed model into multiple simple sub-problems,which makes the problem easier to solve.The experimental results show that the proposed algorithm can suppress noise,preserve edge information,and reduce jagged spectral distortion while ensuring the solution speed.(3)A diffractive spectral image reconstruction algorithm based on low-rank constraints and adaptive spatial-spectral weighted total variation is proposed,which further improves the restoration effect on the basis of the multi-channel spatial-spectral total variation model.This method fully considers the high correlation between spectral images,and introduces a global low-rank constraint,which is used to separate noise-free diffractive spectral images with low-rank characteristics from system noise.In addition,considering the structural information in the spatial horizontal,vertical and spectral directions of each pixel,adaptive weights are constructed to adjust the strength of the local spatial smoothness and local spectral smoothness constraint with each pixel in different regions.The improved model can further suppress the noise and improve the effect of spectral recovery when dealing with the ill-conditioned problem with cross-channel blurs and noise interference.