Block based face recognition approach robust to nose alterations

Neslihan Kose (1) , Nesli Erdogmus , Jean-Luc Dugelay (1)

1 Eurecom [Sophia Antipolis]

Face recognition that is robust to alterations applied on face via plastic surgery or prosthetic make-up can be still considered as a new topic. In this paper, a block based face analysis approach is proposed which provides a fairly good recognition performance together with the advantage of robustness to such kind of alterations. For this study, a simulated nose alteration face database is used which is prepared using FRGC v1.0. Since this is a 3D database, the approach can be tested both in 2D and 3D, which is one of the contributions of this study. Furthermore, differently from previous works, baseline results for the original database are reported. The impact which is purely due to the applied nose alterations is measured using both the proposed approach and the standard techniques which are based on holistic description for comparison. The results indicate that although both 2D and 3D modalities lose precision due to alterations, the proposed approach is superior in terms of both the recognition performance and robustness to alterations compared to standard techniques.

Domaines: Vision par ordinateur et reconnaissance de formes [cs.CV]

Neslihan Kose, Nesli Erdogmus, Jean-Luc Dugelay. Block based face recognition approach robust to nose alterations. IEEE 5th International Conference on Biometrics: Theory, Applications and Systems, BTAS 2012, Sep 2012, Washington DC, United States. pp.1. ⟨hal-00782194⟩