摘要
针对大范围姿态角度问题,提出了一种分级边缘方位场匹配(HEOFM)的人脸特征定位算法。提出了结构Hausdoff距离作为边缘方位场匹配的测度;在此基础上,先根据人脸图像的整体边缘方位场匹配(GEOFM)进行了姿态预估计,得到特征的粗略定位;然后根据特征边缘方位场匹配(FEOFM)在局部区搜索,实现特征的精确定位。实验结果表明:针对小姿态角度人脸图像,与一般的特征检测算法相比,该算法所需训练样本个数较少而定位精度相当;针对中、大姿态角度人脸图像,该算法具有良好的特征定位性能。
In order to improve the facial feature localization under large-scale pose angle, a hierarchical edge orientation field rnatching（HEOFM） based algorithm was proposed. A new measurement coined structural Hausdoff distance was developed for edge orientation field matching. The global edge orientation field matching（GEOFM） was applied to estimate the pose angle and thus obtain the coarse position of the features, and then the further local feature edge orientation field matching （FEOFM） was carded out for precise localization. The experimental results show that, this algorithm achieves almost high localization accuracy with less training examples compared with common detection based localization algorithm under small pose angle,and performs fairly good under moderate or large pose angle.
出处
《光电子．激光》
EI
CAS
CSCD
北大核心
2007年第2期241-244,共4页
Journal of Optoelectronics·Laser
基金
浙江省青年人才培养资助项目(R105341)
