Photometric stereo and appearance capture

Abstract

Photometric stereo recovers surface shape from images taken under different lighting conditions. Auto-calibration photometric stereo methods recover surface shape and lighting directions at the same time. In this thesis, we propose two novel auto-calibration methods. One method exploits special ring-light configurations to resolve general Lambertian surfaces. Another method is based on patch factorization which takes advantage of near-object light sources to resolve ambiguities. For non-Lambertian surface, previous photometric stereo methods often assume parametric or known reflectance model, which limits their application. We propose a new multi-view photometric stereo technique that can work for general isotropic materials. We collect `iso-depth contours' from multiple viewpoints and combine them with multi-view stereo to obtain a precise reconstruction of the complete 3D shape. The spatially varying isotropic bidirectional reflectance distribution function (SVBRDF) is captured by simultaneously inferring a set of basis BRDFs and their mixing weights at each surface point.