Panoramic face mosaicing and recognition using multi-cameras
Fan Yang, Michel Paindavoine, Hervé Abdi, Dominique Arnoult · 2008
applications such as human expression categorization using movement estimation or real-time applications. Also, In this article, we present some development results their cost can easily make these systems prohibitive for rou of a system that performs mosaicing (or mosaicking) of tine applications. In order to avoid using expensive and panoramic faces. Our objective is to study the feasibil time intensive 3D acquisition devices, some face recogniity of panoramic face construction in real-time. To do so, tion systems generate 3D information from stereo-vision. we built a simple acquisition system composed of 5 stan Relatively complex calculations, however, are necessary dard cameras which, together, can take simultaneously 5 in order to perform the auto-calibration and 2D projective views of a face at different angles. Then, we chose an eastransformation[8]. Another possible approach is to derive ily hardware-achievable algorithm, consisting of successive some 3D information from a set of face images, but with linear transformations, in order to compose a panoramic out trying to reconstitute the complete 3D structure of the face from these 5 views. The method has been tested on a face[5][9]. relatively large number of faces. In order to validate our The goal of the present article is to describe a system system of panoramic face mosaicing, we also conducted a which is simple and efficient and also which can potentially preliminary study on panoramic faces recognition, based process 3D faces in real-time. First, we describe our acqui on the principal component method. Experimental results sition system. Then, we describe the method for creating show the feasibility and viability of our system. panoramic face mosaics using successive linear transformations. Next, we present experimental results of panoramic face recognition. Finally we conclude and explore possible 1.