Polarization-Informed Non-Line-of-Sight Imaging on Diffuse Surfaces

Abstract

Non-line-of-sight (NLOS) imaging has relevance in search \\& rescue, medical imaging, remote sensing, and robotics. Although NLOS methods are maturing, NLOS with normal cameras generally requires special occluders in the scene to remove light transport ambiguity. In this paper, it is shown that polarization reveals unique information about occluded environments, and computation in the polarization domain has sparsity benefits that aid the inverse problem. This is demonstrated via non-line-of-sight imaging on rough, everyday surfaces such as office/home walls. If successful, it has the potential to enable direct and indirect occluded light source discrimination and passive shape recovery of hidden objects via shape from polarization.