基于运动检测和小波分析的视频图像人脸检测

Face Detection Based on Movement Detection and Wavelet Analysis within Video Sequence

人脸检测是所有人脸处理系统的基础,而视频中人脸检测问题更具有其特殊的意义。结合运动检测和小波分析技术,提出了一种快速的、具有良好鲁棒性的人脸检测方法。首先基于运动检测得到包含人脸对象的运动区域——人脸图像;然后应用小波分析良好的时频局部化特性和多分辨率特性对运动检测定位的人脸图像进行降维处理,同时得到体现人脸特征的小波系数候选本征脸;最后结合人脸的面部特征及形状分析进一步对结果进行验证。实验结果表明,该方法具有较好的实时性和鲁棒性。

Face detection is the foundation for all the face processing systems, and its applications within video sequence are more significant. A robust and fast face detection algorithm, combined with the movement detection and wavelet technology are introduced. Firstly, locate the movement region which combines the face object based on movement detection, called face image. Following, decompose the face image by utilizing wavelet technology and decrease its dimension, and get the candidate face. Finally, verify the candidate face by the features and shape of face. Experimental results show that this algorithm is mostly real-time and robust.