Computational Depth Imaging Using the Fast Deconvolution Method

Abstract

Pseudo-random spread spectrum photon counting (PSSPC) is a well-established technique for three-dimensional (3D) imaging. Based on the pseudo-random spread spectrum photon counting system, a fast imaging technique that is able to accurately recover multiple depths at individual pixels is presented. Firstly, a pre-filtering algorithm is used to denoise the original data. Then the accelerated Richardson-Lucy iterative deconvolution algorithm is introduced. The method is based on the principles of vector extrapolation and does not require the minimization of a cost function. For multi-depth estimation in the presence of moderate background light, we experimentally demonstrate that our imaging technique outperforms the existing method. We have successfully improved the range resolution from 21cm to 8cm, thus breaking the Full Width at Half-Maximum (FWHM) resolution limit. The separation Root Mean Square Error (RMSE) has been reduced to 3.82cm by the proposed method for the surface-to-surface separation of 8cm. This is a factor of 4 improvement over the conventional method for multi-depth recovery. Also, our imager has achieved 0.5cm lateral resolution by distinguishing two squares closely placed 0.5cm apart from each other.