Abstract

This paper presents a simple yet practical 3-D modeling method for recovering surface shape and reflectance from a set of images. We attach a point light source to a hand-held camera to add a photometric constraint to the multi-view stereo problem. Using the photometric constraint, we simultaneously solve for shape, surface normal, and reflectance. Unlike prior approaches, we formulate the problem using realistic assumptions of a near light source, non-Lambertian surfaces, perspective camera model, and
the presence of ambient lighting. The effectiveness of the proposed method is verified using simulated and real-world scenes.