Photon counting integrated imaging iterative reconstruction method based on Bayesian estimation

Abstract

The invention discloses a photon counting integrated imaging iterative reconstruction method based on Bayesian estimation. A posterior probability model of photon number multi-point estimation is built according to the Poisson distribution in the process of photon counting and the spatial correlation of images. In the process of iteration, a Poisson de-noising algorithm is introduced to update the parameters of the posterior probability model. A photon counting integrated imaging system realizes imaging under the condition that an extremely small number of photons are detected. The quality of target three-dimensional reconstruction is improved.

Description

technical field [0001] The invention belongs to the field of photon counting imaging, in particular to a photon counting integrated imaging iterative reconstruction method based on Bayesian estimation. Background technique [0002] Integrated imaging has attracted extensive attention in the field of 3D imaging because it can record 2D images of different viewing angles of 3D targets and capture information about the direction and intensity of light. unique advantage. Integrated imaging can be realized with the help of microlens array or camera array, while the synthetic aperture integrated imaging system performs multi-channel observation of the 3D scene through the translation of a single camera. Each channel represents the image acquisition of the scene by each camera, and the acquired The two-dimensional image of is called elemental image. [0003] The photon counting imaging system gets rid of the limitations of traditional imaging systems requiring a large number of p...

AI Technical Summary

Problems solved by technology

[0004] In the field of photon counting integrated imaging reconstruction, the Poisson distribution is usually used to establish the photon counting detection model. The traditional method is to use the maximum likelihood estimation (MLE) for the depth reconstruction of integrated imaging. Although the maximum likelihood estimation algorithm It can produce better results, but it lacks prior information of the target object, and it is still insufficient in reconstruction effect. When the ambient photon is low to a certain extent, its application will be limited

Therefore, it is proposed to use the Gamma distribution as the prior distribution of the element image to perform Bayesian estimation reconstruction on the scene. However, the Gamma distribution is not the most appropriate distribution to describe the characteristics of the target, it can only be used as a rough description

Another reconstruction direction is to transform the reconstruction problem of integrated imaging into an inversion problem, construct the imaging system matrix H, and introduce a regularization factor to perform the reconstruction of the penalty likelihood maximum expected value (PMLEM). The condition of the result is the selection of the penalty item in the iterative formula and the uncertainty of parameter adjustment, which leads to the poor adaptability of the scene of the method

Images