A video human face tracker.

This thesis focuses on Automated Face Tracking, which is a major challenge for applications related to intelligent surveillance, law enforcement, telecommunications, human-computer interactions and virtual reality interfaces. The thesis describes a novel strategy for both face tracking and facial feature detection. Face detection is important because it restricts the field of view and thus reduces the amount of computation for further face recognition or transmission, while facial feature detection is important because it enables face normalization and leads to size invariant face recognition. Resembling the way of human perceiving face and facial features, the proposed face tracker initially utilizes motion as the major cues, then human eyes are searched from the area which likely contains a human face. The contributions of this thesis are threefold. First, we have introduced an effective system architecture including a system supervisor controlling five virtual perceptual sensor processes. Second, we have advanced a hybrid approach for face detection and facial feature extraction. An adapted Hough Transform is introduced in search of human eyes within the candidate face region. Finally, implementation issues are explored when using a low-cost Internet camera under a common Windows environment. Methods of using Microsoft Video for Windows (VFW) and real-time programming approaches are employed in this specific face tracking application. Source: Masters Abstracts International, Volume: 40-03, page: 0756. Adviser: H. K. Kwan. Thesis (M.Sc.)--University of Windsor (Canada), 2000