Research And Application Of Image Reconstruction Based On Lensless Encoder Plate Imaging

As an important branch of Computational Optical imaging,lensless coded mask imaging system extends the idea of coded aperture imaging,breaks through many limitations of traditional lens-based imaging system,and has great development potential in the future.At present,Lensless coded mask imaging systems can be divided into three categories according to its imaging principle: lensless multiplexing imaging system,lensless compressive imaging system and lensless microscopic imaging system.In this paper,the first two are studied.The imaging structure,imaging mechanism,coding mask pattern and reconstruction algorithm are analyzed from the optical and mathematical perspectives.At the same time,aiming at the shortcomings of the existing reconstruction algorithm,the improved reconstruction algorithm is proposed,and the superiority of the new algorithm is proved by simulation experiments.The major innovative works are as follows:(1)A total variation regularization reconstruction algorithm based on split Bregman iteration is proposed for reconstruction of lensless multiplexing imaging system.This kind of imaging system mainly uses the method based on Tikhonov regularization to reconstruct image,which performs poorly in reducing noise and preserving edge information.Inspired by the image restoration method based on Bregman iteration,we proposed a total variation regularization reconstruction algorithm based on split Bregman iteration.The proposed algorithm can not only preserve image edge information,but also effectively reduce noise,and has advantages in reconstruction time.The effectiveness of the algorithm is verified by simulation experiments.(2)A reconstruction method based on non-local low-rank and fractional-order total variational mixing regularization is proposed for the image reconstruction of lensless compressive imaging system.This kind of imaging system mainly uses the reconstruction method in the compressed sensing theory to reconstruct the image.Because it compresses the whole scene when sampling the image,the signal dimension is large,so we hope that the sampling rate can be as low as possible.Existing compressed sensing reconstruction algorithms cannot effectively preserve local texture and weak edge information at low sampling rate.To solve this problem,a hybrid reconstruction model based on non-local low rank and fractional order total variation is proposed.By choosing the appropriate fractional order,texture and weak edge are preserved better than other reconstruction algorithms.The experimental results show that the proposed algorithm has better overall reconstruction effect at a lower sampling rate.(3)In order to display the reconstruction effect of different reconstruction algorithms more conveniently and intuitively,an image reconstruction simulation system based on MATLAB platform is designed and developed.The simulation system is divided into two modules: image reconstruction based on lensless multiplexing imaging and image reconstruction based on lensless compressive imaging.The two modules are divided into three sub-modules: data selection module,parameter and algorithm setting module,and reconstruction result display module.In the simulation system,users can select different test images,set different test parameters,test different reconstruction algorithms,and view the results of image reconstruction and performance evaluation.Using this system,the performance of different reconstruction algorithms can be compared and analyzed more conveniently.