文档介绍：International Journal puter Vision 43(3), 167–188, 2001
c 2001 Kluwer Academic Publishers. Manufactured in herlands.
Face Recognition Using the Discrete Cosine Transform
ZIAD M. HAFED AND MARTIN D. LEVINE
Center for Intelligent Machines, McGill University, 3480 University Street, Montreal, Canada H3A 2A7
******@
******@
Abstract. An accurate and robust face recognition system was developed and tested. This system exploits the
feature extraction capabilities of the discrete cosine transform (DCT) and invokes certain normalization techniques
that increase its robustness to variations in facial geometry and illumination. The method was tested on a variety of
available face databases, including one collected at McGill University. The system was shown to perform very well
pared to other approaches.
Keywords: Face recognition, discrete cosine transform, Karhunen-Loeve transform, geometric normalization,
illumination normalization, feature extraction, pression
1. Introduction This paper discusses a putational approach
to face recognition that, bined with proper
Face recognition by humans is a high level visual face localization techniques, has proved to be very
task for which it has been extremely difficult to con- efficacious.
struct detailed neurophysiological and psychophysi- This section begins with a survey of the face recog-
cal models. This is because faces plex natu- nition research performed to date. The proposed ap-
ral stimuli that differ dramatically from the artificially proach is then presented along with its objectives and
constructed data often used in both human - the motivations for choosing it. The section concludes
puter vision research. Thus, developing puta- with an overview of the structure of the paper.
tional approach to face recognition can prove to be
very difficult indeed. In fact, despite the many rel-
atively essful attempts to puter- . Background and Related Work
based face recognitio