Cue Combination for Robust Real-Time Multiple Face Detection at Different Resolutions

The face detection problem, defined as: to determine any face -if anyin the image returning the location and extent of each [Yang et al., 2002], seems to be solved, according to some recent works [Schneiderman and Kanade, 2000] [Viola and Jones, 2001]. Particularly for video stream processing, these approaches focus the problem in a monolithic fashion, forgetting elements that the human system employs: temporal and contextual information, and cue combination. The work summarized in this abstract describes an approach for robust realtime multiple face detection which combines different cues. The resulting approach achieves better detection rates for video stream processing and cheaper processing costs than outstanding and public available face detection systems.