Research On 3D Reconstruction Technology Of Photon Counting Integrated Imaging

Photon counting integral imaging(?)system can reconstruct the three dimensional(3D)objects passively under photon-starved conditions by recording different perspectives of the scene with a photon counting array.Therefore,photon counting ? system,as a 3D passive imaging system,can be widely applied in the fields of military defense and national economy.Traditional photon counting ? reconstruction algorithms either select the gamma distribution as a priori under the framework of Bayesian or transform the reconstruction problem into an inverse problem by introducing a regularization operator in an iterative formula.These methods rely heavily on the choice of prior distribution and have the uncertainty of parameter adjustment,resulting in instable performance and poor adaptability for different scenes.In order to make up for these shortcomings,this paper makes full use of the similarity across the view in the photon counting ? system and new methods for its 3D reconstruction are proposed.Firstly,a new method of photon counting ? reconstruction based on adaptive Bayesian estimation is proposed.A local adaptive mean factor is introduced to help the estimation of Bayesian posterior probability.The experimental results show that the depth slice images reconstructed by the proposed method have richer gray level and higher peak signal to noise ratio than those reconstructed by traditional Bayesian method.Secondly,photon counting ? using compound photon counting model is proposed.Through maximizing a likelihood function with pixel-based adaptive information derived from the additivity of Poisson distribution,variance stabilizing transformation combined with block-matching and 3D filtering algorithm is also applied to reconstruct the 3D depth slice images of higher quality at low light level.For both proposed algorithms,the expected number of photons and neighborhood range are analyzed theoretically and experimentally.Both proposed algorithms are suitable for the reconstruction of photon-starved environments,and effective expansion of neighborhood range will further improve the reconstructed image.Finally,photon-limited images are captured by Electron Multiplying CCD camera in case of a few photons illumination.Integral imaging 3D reconstruction based on Gaussian detection model and Poisson detection model are respectively derived and better performance by the proposed adaptive Bayesian algorithm is proved through comparative experiments.