Real-time human face tracking

Pan, Wenbo (2000) Real-time human face tracking. Masters thesis, Memorial University of Newfoundland.

[img] [English] PDF - Accepted Version
Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission.

Download (12Mb)

Abstract

A real-time human face tracking system has been developed at the Multimedia Communication Laboratory of Memorial University of Newfoundland to investigate fast, efficient, reliable and flexible face finding and tracking techniques. By subtracting a well-maintained background from an incoming image, an object and background segmentation map can be constructed. The foreground object is outlined by a "draping" operation on the segmentation map. Once the drape is settled, an innovative head identification method consisting of exhausting head searching followed by head merging achieves accurate head extraction and identification. In order to tackle the problems associated with variations in lighting, local and global background movements and shadows in the background scenes, a multi-state background self-generating/adjusting method is applied. This allows the system to switch automatically between background formation and simple face tracking. The draping is applied on the inter-frame variance of incoming images to identify moving areas and thus to generate the background. Median filtering, multiple direction draping and polynomial interpolation are developed and incorporated into this system to overcome the possible pitfalls in the resultant drape. The background is updated automatically in real time once the changes in the background are detected to exceed a given threshold. Experiments show that the new real-time system is a robust and effective tool for extracting human heads from a very complex non-stationary background.

Item Type: Thesis (Masters)
URI: http://research.library.mun.ca/id/eprint/9635
Item ID: 9635
Additional Information: Bibliography: leaves 69-72.
Department(s): Science, Faculty of > Computer Science
Date: 2000
Date Type: Submission
Library of Congress Subject Heading: Computer vision; Pattern recognition systems

Actions (login required)

View Item View Item

Downloads

Downloads per month over the past year

View more statistics