Research On Reconstruction Method And Experiment Of Low Light Level Image

Low-light-level(LLL)refers to a dark environment with a small amount of natural light(less than10^-3 lx).As an extension of image reconstruction technology to the direction of low illumination,LLL image reconstruction can effectively make up for the defect that human eye can not effectively identify the target scene in LLL environments such as night.It has an extremely important application value for deep sea exploration,military operations,medical detection,etc.The work of this paper is carried out in a LLL environment,an imaging experiment platform is built to obtain LLL images,and the restoration and reconstruction methods of LLL images are studied as follows:1.In order to get images of 3D target objects in the LLL environment,an imaging system is designed according to the principle of off-axis integrated imaging.In terms of hardware,the multi-pixel photon counter(MPPC)is combined with the optical imaging structure,the optical path is designed,and a LLL integrated imaging experimental platform is built.In terms of software,a system control program is written to collect and record image information.It is realized the imaging of 3D target objects in a low illumination environment of10^-3-1 0^-6 lx.2.In the research of LLL image restoration,a Bayesian regularization photon counting image restoration algorithm is proposed to solve the problem of image quality degradation caused by pollution such as background noise.The algorithm estimates the photon count value of LLL images based on the Poisson distribution characteristics of photon count,and introduces Bayesian estimation into the regularization algorithm to establish the objective function.In the process of solving the objective function,the prior condition is set as the Gamma distribution of the expected value of photon counting,and the error part is expressed as the form of norm.Regularization parameter is iteratively solved to find the optimal interval,so as to estimate the optimal value of the photon count.Analysis with advanced algorithms such as BM3D shows that the proposed method can effectively restore and improve the quality of LLL images.3.In the research of LLL image reconstruction,based on the SIFT algorithm,a hybrid feature point detection and matching method is proposed.This method combines FAST corner detection on the basis of SITF feature descriptors,which improves the detection speed and increases the number of recognition feature points.In the matching process,SIFT matching improves the accuracy of matching feature points.When removing mismatches,the probability of removing mismatched points by the RANSAC algorithm reaches 80%,which greatly reduces mismatched points.This method solves the problem that the LLL image is extremely lack of photon information and the feature details are not obvious,so that the corner points cannot be accurately identified.Through comparative analysis,the LLL image reconstructed by this method has more detailed information and higher recognition degree.The LLL image restoration algorithm and feature point detection and matching method proposed in this paper can effectively improve the quality of LLL image reconstruction.Compared with other algorithms,the peak signal-to-noise ratio,mean square error and correlation coefficient of LLL image restoration are all improved by 3%-7%,and the reconstructed gray average gradient,Laplacian operator,and contrast are increased by 10.4%,32.3%,and 2.9% respectively.