摘要
提出一种基于离散小波变换 (DWT)的人脸检测与特征定位算法 ,算法针对人脸的不同特征 ,选择小波变换相应最有效的频段和分量 .运用椭圆检测算法、模板匹配算法、Hough变换等方法 ,并结合人脸各特征间的几何特征关系 ,采用由粗到细的方法分步地检测人脸范围 ,分割定位人脸的眼睛、虹膜、嘴、鼻尖等特征 .该算法对人脸姿态、表情变化和遮盖物具有一定的鲁棒性 ,同时具有计算量小、效率高的特点 .
A new method of facial features localization based on discrete wavelet transform ation (DWT) is presented. According to different facial features, different comp onents and spectrums from wavelet transformation are selected. Then face area an d facial features such as eyes, irises, mouth and nose, are localized step by st ep by using the geometric features among facial features, ellipse detection algo rithm, template matching algorithm and Hough transformation. This method is more robust for changes of facial pose, expression and covering, meanwhile it needs less computation.
出处
《东南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2004年第B11期117-120,共4页
Journal of Southeast University：Natural Science Edition
关键词
人脸特征定位
小波变换
椭圆检测
模板匹配
HOUGH变换
facial features localization
wavelet transformation
ellipse detection
templat e matching
Hough transformation
