A Robust Particle Filter-Based Face Tracker Using Combination of Color and Geometric Information

摘要

Particle filtering is one of the most successful approaches for visual tracking. However, so far, most particle-filter trackers are limited to a single cue. This can be a serious limitation, since it can reduce the tracker's robustness. In the current work, we present a multiple cue integration approach applied for face tracking, based on color and geometric properties. We tested it over several video sequences and we show it is very robust against changes in face appearance, scale and pose. Moreover, our technique is proposed as a contextual information for human presence detection.