Three-dimensional photon counting integral imaging using Bayesian estimation

Abstract

We propose a new estimation method for 3D object reconstruction using photon-counting integral imaging. Earlier studies used maximum likelihood estimation (MLE) as a classical statistical method to reconstruct 3D images from photon-counting elemental images. We use an alternative statistical method known as the Bayesian method, which is more flexible and may perform better than MLE in terms of the mean square error (MSE) metric. The performance of the new reconstruction method is illustrated and compared with MLE by using the MSE. To the best of our knowledge, this is the first report to use the Bayesian method for 3D reconstruction of photon-counting integral imaging.

Polarimetric 3D integral imaging in photon-starved conditions

“…Note that MSSIM values for DoP are extremely low. Different methods have been proposed for improving photon-count integral imaging reconstruction using statistical approaches [24][25][26]. In this paper, we use total variation denoising.…”

“…Integral imaging under extremely low light level scenes using photon-counting elemental images has been shown as a particularly interesting topic. Several papers have demonstrated the feasibility of 3D image visualization using a low number of photons [23][24][25][26]. Moreover, polarization and 3D data are closely related: for instance, 3D information can be derived from the analysis of the state of polarization of the light reflected by the analyzed object [4,27].…”

Estimation of the degree of polarization in low-light 3D integral imaging

“…Several techniques have been proposed for producing an InIm from elemental images recorded with few photons [11][12][13][14]. In particular, the Maximum Likelihood Estimation (MLE) is a simple but effective method that produces reasonable results.…”

Efficient reconstruction of 3D images from photon starved integral imaging using PMLEM

“…Integral imaging [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18] (InI) is a technique that uses lenslet arrays for capturing and displaying three-dimension (3D) images. It was originally proposed by Lippman 2 in 1908 and it was revived recently owing to the availability of high performance capturing and displaying systems.…”

“…Integral imaging may be viewed as a multichannel system 1 , each channel generating an elemental image. As such there is large redundancy in the captured data, thus permitting imaging of obscured objects 3 , 3D target recognition 1 and 3-D image restoration in photon starved conditions [14][15][16][17][18] .Several approaches for imaging in photon starved environments and photon counting imaging have been introduced [14][15][16][17][18] . Using computational reconstruction, the captured data can be digitally processed to improve the visual information 1,4,12 .…”