A unified approach to video face detection, tracking and recognition
Simon Clippingdale, Takahiro Ito · 2003
We describe a prototype system which recognizes faces in video by tracking variable parameters such as face pose, size etc., and integrating information about identity over time. The detection, tracking and recognition processes are unified, using multi-resolution deformable template matching in a hypothesis generation and revision framework. We discuss test results and various possibilities for improving performance.