Plug-and-Play for Joint Deblurring and Guided Super-Resolution of Single-Photon 3D Lidar data

This paper presents a new method for joint deblurring and super-resolution of 3D single-photon Lidar images. Adopting a plug-and-play framework, the method alternates between a data fidelity iterate, and a guided filtering (GF) step which can be performed by any existing GF algorithm. The resulting analytical updates are efficient and are easily adapted to different upscaling factors or arbitrary blur kernels. Thanks to the GF algorithm, the algorithm shows good denoising performance when imaging in extreme conditions leading to high background noise. Experiments on simulated and real data demonstrate the good performance of the proposed strategy in terms of depth maps deblurring and super-resolution in presence of high levels of noise.