Deep Photometric Stereo Network

Abstract

This paper presents a photometric stereo method based on deep learning. One of the major difficulties in photometric stereo is designing can appropriate reflectance model that is both capable of representing real-world reflectances and computationally tractable in terms of deriving surface normal. Unlike previous photometric stereo methods that rely on a simplified parametric image formation model, such as the Lambert's model, the proposed method aims at establishing a flexible mapping between complex reflectance observations and surface normal by the use of a deep neural network. As a result we propose a deep photometric stereo network (DPSN) that takes reflectance observations under varying light directions and infers the corresponding surface normal per pixel. To make the DPSN applicable to real-world objects, a database of measured bidirectional reflectance distribution functions (MERL BRDF database) has been used for training the network. Evaluation using simulation and real-world scenes shows effectiveness of the proposed approach over previous techniques.