基于三维可变形模板的眼睛特征提取  

EYE FEATURE EXTRACTION BASED ON 3D DEFORMABLE TEMPLATE

作　　者：黄万军[1] 尹宝才[1] 陈通波[1] 孔德慧[1] 

机构地区：[1]北京工业大学计算机学院,北京100022

出　　处：《计算机研究与发展》2002年第4期495-501,共7页Journal of Computer Research and Development

基　　金：国家"八六三"高技术研究发展计划基金(863-306-ZD03-01-2);北京市科技新星计划基金资助

摘　　要：眼睛特征提取在人脸感知的各种应用中有着非常重要的作用．为了解决人脸垂直旋转角度比较大时．眼睛特征提取的问题，提出了一种新的基于三维可变形模板的眼睛特征提取方法．此方法采用了两个新提出的技术，一个是人脸姿态估计策略用于估测人脸的垂直旋转方向，另一个就是三维可变形模板匹配技术用于具体提取眼睛的精确特征．实验表明该方法能够适应人脸图像垂直旋转角度的变化，获得很好的眼睛特征提取结果．Eye feature extraction has become one of the important issues in the applications of facial image perception. A new approach for eye feature extraction based on 3D deformable template has been proposed to extract the exact eye features of the facial images with large vertical rotation angle. The approach includes two new techniques. One is pose estimation strategy that can be used to obtain the vertical rotation angle of the facial image, and the other is 3D deformable template matching which can be employed to extract the eye features of the facial image. The experimental results demonstrate the efficiency of the approach for both frontal facial image and the facial image with large vertical rotation angle.

关 键 词：三维可变形模板 眼睛特征提取 模板匹配 姿态估计 能量函数 人脸识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]