Near-infrared image based face recognition: Approach and system

We present a novel, near infrared (NIR) image based, approach and system for face recognition. There are two novelties which makes it better than existing methods. The first is an imaging hardware system for acquisition of face images that are stable to environmental illumination, subject only to a monotonic transform in the gray tone. The second novelty is a novel algorithm for illumination invariant face recognition: Local binary pattern (LBP) is used to compensate for the monotonic transform, making a truly illumination invariant face representation; an AdaBoost learning procedure then selects the best subset of LBP features and builds a highly accurate face recognition engine. Highly accurate and fast face recognition systems can be built using the hardware and algorithm. Results of both technology evaluation and scenario evaluation are provided to demonstrate the power of the present approach.