RFIA 2012, Reconnaissance des Formes et l'Intelligence Artificielle, 24-27 Janvier 2012, Lyon, France
In this paper, we propose to fuse different 3D-3D matchers to achieve high effectiveness in distinguishing facial differences between distinct individuals even in presence of facial expressions. We present four different matchers: (E1) Elastic radial curves, (E2) MS-eLBP: multi-scale extended LBP, (E3) TPS algorithm, in addition to one baseline matcher (Eref) ICP algorithm for rigid registration. Evaluations conducted on FRGCv2 (Face Recognition Grand Challenge) dataset on individual matchers and on their fusion (Multiple matching approash) show interest of the proposed approach. It achieves 98.6% as rank-1 recognition rate in presence of facial expressions which is competitor compared to state-of-the-art results.
Type:
Conférence
City:
Lyon
Date:
2012-01-24
Department:
Sécurité numérique
Eurecom Ref:
3590
Copyright:
© EURECOM. Personal use of this material is permitted. The definitive version of this paper was published in RFIA 2012, Reconnaissance des Formes et l'Intelligence Artificielle, 24-27 Janvier 2012, Lyon, France and is available at :
See also:
