Abstract

The computer identification of human faces is a fundamental element of high level human-computer interaction. It must be both accurate and easy to use. In order to realize fully automatic facial image processing and face identification based on 2D intensity images, it is necessary to accurately extract the standard face view from the input face image automatically. If the standard face view is not extracted correctly, identification accuracy remains poor. Therefore, manual extraction of the standard face view is used in most previous facial image processing research.

© 1992 Optical Society of America