Circular-polarization-assisted 3D shape reconstruction with snapshot full-Stokes imaging

The polarization signature of reflected light has been used for contactless 3D shape reconstruction of objects. Conventional Shape from Polarization methods employ unpolarized light to illuminate the tested surface and remain constrained to the linear regime by analyzing AoLP and DoLP to compute the surface’s normals and corresponding 3D height map. In this paper we expand the measurement scheme to the full-Stokes regime by illuminating with circularly polarized light and incorporating the S 3 information in the analysis, enabling the computation of DoCP (in addition to AoLP) with a proprietary full-Stokes camera to provide the corresponding surface’s normals under specular or diffuse reflectance models. We compare both methods experimentally by measuring objects of different geometries and evaluating the 3D reconstructions obtained. Results show that working with circularly polarized illumination and full-Stokes detection enriches the polarization information available for shape reconstruction, being most beneficial when conventional DoLP-based 3D reconstruction produces strong flattening or ridge-like artifacts, while for smoother natural objects the improvement may be more limited.