Design and Implementation of a Bimodal Face Recognition System

Abstract

Visible light face images usually have a high resolution; however, the performance of face recognition using visible light face images is usually affected by the varying illumination. It seems that near infrared face recognition might be little influenced by the varying illumination, whereas near infrared face images usually have a low resolution and some facial marks such as scars and moles cannot be reflected by the image. In this paper, we develop a low-cost bimodal face recognition system. The system first captures the visible light and near infrared images of the face and then integrates them for face recognition. The paper also proposes a score level fusion method to combine visible light and near infrared face images for face identification. The experimental results show that the proposed method performs very well in bimodal face recognition. Moreover, the paper provides a true bimodal face image database, in which the visible light and near infrared face images are captured simultaneously.