摘要
简述了近年来主要的人脸跟踪方法,提出一种基于帧间差分与静态特征相结合的人脸跟踪算法。算法利用视频图像的运动信息做帧间差分确定人脸搜索的矩形区域,利用肤色模型提取肤色区域,通过积分投影确定人脸的椭圆范围,并对矩形范围内图像做边缘提取、膨胀等处理,通过水平积分投影获得人脸特征曲线,并通过特征曲线验证椭圆范围内是否存在人脸,实现人脸自动跟踪。
The main methods of face tracking were introduced, and a new method based on both inter-frame differential and static character was put forward. The method makes inter-frame different using the movement information of video-image and obtains the face＇s approximately rectangle district and pick-up the complexion area with the complexion model, and distilsl the face-edge in the ~ctangle area. Then, the image was expanded and the level-projection was made. The character curve was used to validate the existing face in the complexion area. Consequently, automatically face tracking is realized.
出处
《苏州科技学院学报（工程技术版）》
CAS
2006年第2期74-78,共5页
Journal of Suzhou University of Science and Technology （Engineering and Technology）
关键词
帧间差分
肤色模型
特征曲线
人脸跟踪
inter-frame differential
complexion model
character curve
face tracking
