Face detection for image annotation

The problem of face detection in images - locating image areas correspondin g to human faces - has received a considerable amount of attention in recen t years due to numerous possible applications. In this paper, we describe a face detection system for an image annotation task that requires detection of faces at multiple scales against a complex background. The proposed sys tem is a top-down system; it uses the presence of skin tone pixels coupled with shape and face-specific features to locate faces in images. The main d istinguishing feature of the proposed system is its use of an iterative reg ion partitioning procedure to generate candidate face regions. To date, the performance of the system has been rested with over 200 test images of var ying complexity with promising results. (C) 1999 Elsevier Science B.V. All rights reserved.