Surface shape estimation from photometric images

One of the most studied techniques for recovering surface shapes using a computer vision system is the photometric. This method is based on the analysis of one or several images of an object illuminated from a known direction. This kind of images can be considered reflectance maps that give us information of the surface gradient. In computer vision, the problem of recovering 3D information from shaded images is considered an inverse problem. To integrate surface gradient information it is proposed a regularization technique that gives a stable solution of the inverse problem and allows the possibility of reducing errors caused by noise. Results applied on synthetic and real experimental images are presented. (C) 2004 Elsevier Ltd. All rights reserved.