In Search of an Optimal Set of Images for Uncalibrated Photometric Stereo

Abstract

In this paper, we draw on the ideas of previous work done by Drbohlav and Chantler [5], where a theoretical study on the optimal lighting configuration for photometric stereo was developed. They demonstrated that the known n light source azimuth angles should be equally spaced over 360/n degrees and the n light source elevation angles should be constant. In this paper, our aim is to develop a method to find a set of images that best satisfies this lighting configuration, assuming that light source directions are unknown. To this end, we estimate lighting coefficients using a harmonic spherical illumination model and a reference surface. Once the set of light source directions is estimated, we select those images that best satisfy the optimal lighting configuration. Experiments on face datasets show that the selected images are the most suitable for the purposes of calculation of height maps through surface integration.