Depth recovery in time of flight range sensors via compressed sensing algorithm

Abstract

Stereo vision, structured light, and time of flight (ToF) are different range imaging techniques that acquire a scene and provide different features such as a depth map and an amplitude image. Compared to other imaging techniques, ToF can measure depth with high speed and good precision according to state of the art. However, it faces multipath interference (MPI) problems that give rise to an error in radiance information. Exploiting the sparsity of the received signal, we solved the multipath interference problem with the help of compressed sensing sparse recovery algorithms with some modification such as applying positivity constraint and proximity constraint. The modification in the algorithm has increased its robustness and proved to be successful in detecting the interference up to two paths successfully. We validated the approach by providing experimental results on synthetic data with ground truth that demonstrated its efficiency and accuracy to give MPI free output. Moreover, we applied a modified sparse recovery algorithm to real data and compared the result with the state-of-the-art methods. It shows better performance in separating the direct path from the multipath component with high accuracy.