A FACE IMAGE RECOGNITION SCHEME WITH STRONG TOLERANCES TO LIGHTING FLUCTUATIONS (TP-P2)

Author(s) : Kenji Matsuo (KDDI R&D Laboratories Inc., Japan) Masayuki Hashimoto (KDDI R&D Laboratories Inc., Japan) Atsushi Koike (KDDI R&D Laboratories Inc., Japan)

Abstract : In this paper, we propose a face image recognition scheme that has strong tolerances to changes of lighting environment. The current spread of cellular phones with camera device made biometrics certification of face image possible, without adding extra devices. However, face images taken by cellular phone are fluctuated by lighting environments, because cellular phone would be used wherever indoor and outdoor. Therefore, this system has the most important problem that lighting fluctuations decreases its recognition accuracy. Our proposed scheme is very simple. This scheme uses lighting canonical space and creates a virtual subspace which contains various elements of lighting, without actually taking many face images under different environments. Simulation results show that proposed scheme can achieve not only lower equal-error-rate of verification but also higher precision of identification than conventional subspace scheme.

Menu