Depth Estimation Through a Disparity-Based Photometric Stereo

Abstract. Here we revisit a recently introduced process of relative depth estimation through the matching of photometric stereo images. By considering the general solution of the differential equation which relates surface depth to the disparity map produced by the matching process, we are able to obtain a more consistent formulation of such disparity-based approach to photometric stereo. Also, we introduce a multiscale matching procedure, based on a new stochastic metaheuristic for combinatorial optimization, which yields more reliable disparity maps in shorter processing times. Finally, we employ a simple least-squares regression in a calibration strategy for estimating the parameters required by our reconstruction approach. Shape-estimation experiments with real images are presented.