4. Real-Time Tracking of Faces and Facial Features of Multiple People

The problem of tracking the face and other body parts in image sequences has become a popular area of research due to emerging applications in human computer interface, surveillance systems, computer games, model-based video coding, etc. The task of face detection and facial feature motion analysis represents an interesting, challenging problem in computer vision. These algorithms are just the starting point of other aspects of facial analysis, but they provide a machine with a sense of awareness of the user's presence, activity, and expression.

We have implemented a fully automatic, person-independent, real-time system capable of detecting and tracking multiple faces and nine facial features. These facial features are the outer eye corners, eyebrow corners, nostril and mouth corners. The core of this system consists of a set of information-based face and facial feature detectors.

Click on the following images to download video files (MPEG format, aprox. 2 MBytes) that illustrated the result of the face and facial feature tracker.