Generalisation of Photometric Stereo technique to strongQstrong.pdfVIP

Generalisation of Photometric Stereo technique to Q-illuminants Vasileios Argyriou, Svetlana Barsky and Maria Petrou Imperial College London v.argyriou@, s.barsky@, maria.petrou@imperial.ac.uk Abstract We present a generalisation of the 4-light photometric stereo technique to an arbitrary number Q of illuminants. We assume that the surface reflectance can be approximated by the Lambertian model plus a reflectance component. The algorithm works in a recursive manner eliminating the pixel intensities affected by shadows or highlights based on a least squares error technique, re- taining only the information coming from illumination directions that can be used for photometric stereo reconstruction of the normal of the correspond- ing surface patch. We report results for both simulated and real surfaces and compare them with the results of other state of the art photometric stereo algorithms. 1 Introduction Photometric stereo (PS) methods can derive orientation and reflectance values for each pixel, using multiple images captured from the same view point under different illumina- tion directions [18]. In order to recover gradient information of a surface patch, photo- metric stereo assumes that neither highlights nor shadows are present in the images used. In fact under orthographic projection, the only light source direction illuminating a scene which can result in an entirely shadowless image is the one aligned with the viewing di- rection. Therefore, shadowing is nearly unavoidable in most images. Most reconstruction algorithms either ignore the problematic pixels [3] or explicitly exclude data with shadows and highlights [12, 14]. Since at least three images, corresponding to three different lighting directions, are needed to