Automatic Human Face Detection for a Distributed Video Security System
Fei Zuo · 2002
Abstract—This paper ∗ presents a home video security system architecture, which employs pseudo real-time human face detection and recognition techniques for automatic person identification. In this system architecture, the individual processing steps, e.g. face detection and recognition, take place at different locations in the home network. The facial information conveyed in the network is encoded using MPEG-4 coding techniques to enable low bit-rate transmission at low cost within a home environment. In this paper, we primarily focus on the face detection part of the system. We use a combination of coding and analyzing techniques giving direct facial properties to perform face detection. The processing is divided into the following steps: (1) skin-color segmentation, (2) histogram analysis for facial feature detection, and (3) probability-based confidence value evaluation of facial features. The experimental results show that our approach of face detection offers near real-time processing speed with good detection capability and sufficiently reduces the video transmission cost over the network.