Research On Photon Counting Imaging Method Based On Prior Knowledge Constraints

Single photon counting imaging is a novel computational imaging technology,which counts every photon collected by reflected light,and can detect the target under extremely weak light conditions.In contrast,other imaging technologies will encounter difficulties in capturing effective data under this illumination condition.Therefore,this technique has attracted considerable interest in many fields,including space monitoring,biological imaging,fluorescence imaging,and microscopic imaging.Using photon counting imaging system can effectively improve the detection sensitivity,but the inherent background noise in the imaging system and the dark count noise of single photon detector will affect the final image quality.In order to suppress the influence of noise in the detection process,the detector often needs to capture thousands of photons in each pixel to achieve a more accurate reflectivity image.However,under low light conditions,the actual photon count is very low.How to capture high-quality images reflecting the characteristics of the target under the condition of limiting the number of photons has become a key problem in this field.Aiming at this challenge,this paper has carried out theoretical and experimental research on photon counting imaging method based on prior knowledge constraint.The main work and innovations are listed as follows:1.A robust single photon imaging method based on spatial correlation and total variation sparse regularization constraints has been studied.Firstly,in order to reduce the influence of the non-uniformity of the background count,a robust Poisson negative log-likelihood function is proposed as the data fidelity term.The background count is constituted of a constant representing the average background count and a sparse variable denoting the background count of some pixels deviated to the average value a lot.Then the total variation is incorporated as one prior constraint term to reduce the influence of noise whilst preserving edges.Finally,to reduce the truncation error,we added another constraint based on the counting information from spatially correlated points rather than a single point.This constraint will be helpful in increasing reflectivity levels in single-photon counting imaging.2.An image enhancement method based on polarization prior for single photon counting has been studied.By observing that the signal-to-noise ratio of special polarization single photon counting image is higher than that of traditional single photon counting image,we can group the similar blocks of polarization single photon counting image to form an "anti-noise" dictionary with high signal-to-noise ratio.On this basis,a non-local prior sparse representation constraint based on "anti-noise" dictionary is constructed.By introducing the negative Poisson log-likelihood function to maintain the data fidelity,and introducing the total variation constraint to reduce the noise,a complete single photon counting image denoising model is established.3.A super-resolution reconstruction method of single photon counting image has been studied.We divide the high-resolution image into three parts,including the known initial low-resolution image,the missing pixels in the oblique direction and the missing pixels in the vertical direction.According to the nonlocal similarity of the image,the points to be estimated are modeled as a linear weighted combination of their local and nonlocal neighbors to achieve image restoration.We combine super segmentation and denoising to build a complete reconstruction model,which can solve the problems of super segmentation and denoising simultaneously in the field of photon counting imaging.