Research On Reconstruction And Denoising Method Of Fast Correlated Imaging Target

Correlation imaging,also known as ghost imaging,is a new imaging method derived from quantum theory.Traditional optical imaging records the information of an object directly with a detector.However,ghost imaging independently detects the light intensity information and light field distribution information of the light field through two or more detectors,and then performs correlation calculation to reconstruct the image of the target object.Traditional correlation optical imaging schemes require a large amount of computation and long imaging time in order to obtain the target reconstructed image with higher quality.Therefore,this paper aims to shorten the ghost imaging time and improve the quality of ghost imaging.Based on the traditional correlation optical imaging theory and the interpolation algorithm,this paper proposed an interpolation correspondence differential ghost imaging scheme.Firstly,we downsampled the target transmission(reflection)information and compressed the amount of information involved in the correlation reconstruction.The average of multiple iterations of the signal was used as the judgment threshold for logic selection.The light field distribution information corresponding to the light intensity larger and smaller than the threshold is screened out in the reference light path respectively,and then statistically averaged the corresponding light field distribution information to reconstruct the image of the target object.Finally,the reconstructed image was restored to the original pixel size by using interpolation algorithm.In this paper,several typical images of grayscale with different features were used as simulation samples.The experimental results showed that the interpolation correspondence differential ghost imaging proposed in this paper generally improved the peak signal-to-noise ratio by 15% to 20%,the structural similarity by 20% to 130%,and reduced the reconstruction data by 87.5%,compared with the traditional ghost imaging method.Improving the reconstruction algorithm from the perspective of ghost imaging experiments not only shortened the computation time but also improved the quality of the reconstructed image.In this paper,noise removal of reconstruction results of ghost imaging was studied.Four image denoising algorithms,such as mean filtering,bilateral filtering,convolutional neural network and three-dimensional block matching,were used to remove the noise of common images and correlation reconstruction images.It was found that the convolutional neural network denoising algorithm had the best effect on the noise removal of common images,while the three-dimensional block matching method had the best effect on the correlation reconstruction images.The denoising results of the three-dimensional block matching method were compared with the original correlation reconstruction images.The average increase of peak signal-to-noise ratio and structural similarity was about 32.56% and 183.43%,respectively.