中文摘要：

研究了基于滑动窗口的视频实时人脸检测，提出了刚性运动估计（RME）算法。该算法以小尺度人脸瞬时刚性运动为假设，根据几何变换对窗口图像的运动进行描述，以光流代替运动矢量计算运动参数进而识别窗口图像的刚性、非刚性运动类型，通过排除非刚性窗口以提高人脸检测效率。对比实验与分析表明，该算法在准确率与时间效率方面具有优势。

英文摘要：

The real-time video face detection based on sliding motion estimation （RME） was proposed. The algorithm window was studied, and a detection algorithm using rigid describes the motion of window images according to geometric transformation on the assumption that the instantaneous motion of a face in small scale is rigid, and under this condition, the motion types of rigid and non-rigid are recognized by substituting the motion vector with the optical flow to calculate motion parameters. On this account, the efficiency of face detection is promoted by eliminating non-rigid windows. The comparative experiment and analysis show that this algorithm has the advantages in accuracy rate and temporal efficiency.