摘要
介绍了一个完备、新颖的彩色序列图像人脸检测系统.该系统利用颜色信息和运动信息发现人脸可能出现的区域,然后利用人脸的几何特征进行验证,并对正确检测到的人脸提取其面部特征进行精确定位.实验结果表明,该系统具有较强的抗光照强弱变化、人脸侧转、旋转等干扰因素的能力,并能根据输入图像的质量选择输出结果的定位精度,在提高定位精度的同时确保了系统的可信度.
This paper introduced a new integral face detecting system in color image series, which is a prior part of an auto face recognizing system. The system uses color information and motion information to find possible face regions. Then by checking the shapes of the regions, it can prove the true faces. At the final step, it uses face features to locate faces more precisely. The proposed system is robust under many conditions, such as changes of the intensity of illumination, rotations of faces and changes of the number of faces. And the system can choose the precision of its detecting result according to the quality of the input image series in order to increase the precision without the decrease of the reliability of the system.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
2003年第9期1438-1442,共5页
Journal of Shanghai Jiaotong University
