Multispectral Face Recognition

Abstract

Face recognition technology has steadily progressed from adequately handling only well-controlled imagery to tackling increasingly more realistic conditions. This progression has seen the introduction of nuisance factors such as pose, illumination, occlusion and facial expression as integral components of the standard face recognition problem. A large body of research has accrued, aimed at coping with increased levels of image variability while maintaining high recognition performance. Variation in level and nature of illumination is among the most insidious problems for recognition algorithms, and thus a considerable portion of that research centers around it. Among other tech- niques, the use of thermal infrared imagery, by itself or in combination with other modalities, has been proposed as an alternative means of handling the problem of variable illumination conditions.