Prolegomena for Robust Face Tracking

Real-time 3D face tracking is a task with applications to animation, video teleconferencing, speechreading, and accessibility. In spite of advances in hardware and efficient vision algorithms, robust face tracking remains elusive for all of the reasons which make computer vision difficult: variations in illumination, pose, expression, and visibility complicate the tracking process, especially under real-time constraints. This paper considers the problem of robust face tracking. First, we survey recent work in face tracking, which helps us determine what is required for robust face tracking. In particular, we note that robust systems tend to possess some state-based architecture comprising heterogeneous algorithms, and that robust recovery from tracking failure requires several other facial image analysis tasks. Finally, we brie y describe one system for robust face tracking that uses Incremental Focus of Attention (IFA), a state-based architecture which allows fast recovery of lost targets within a unified framework.