Abstract: This paper proposes a fast face detection method in real-time video. It uses the DirectShow technology to grab the video frame, which has significantly increased system speed. After some pre-processes, the system detects the skin color by the statistical property of human skin to make the skin areas as the regions of interests. And these regions of interests are scanned by this improved AdaBoost algorithm for the further detection. Experimental results proves the validity of this method.

Key words: face detection, DirectShow, skin color segmentation, AdaBoost algorithm, partial scan

摘要： 提出一种实时视频中的快速人脸检测方法。采用DirectShow获取和回放视频帧，进行前处理。对所得彩色图像进行肤色分割，通过肤色的统计特性得到人脸候选区域，将该区域设置为感兴趣区域。运用基于局部扫描的AdaBoost算法对每个感兴趣区域进行检测。实验结果证明了该方法的有效性。

关键词: 人脸检测, DirectShow技术, 肤色分割, AdaBoost算法, 局部扫描

CLC Number: 

• TP391