摘要
研究一种在复杂背景下带一定角度姿势旋转的彩色人眼检测算法,解决了长期以来光照条件不均匀和人脸的伸缩旋转对检测过程中的干扰问题。算法在颜色空间中结合Mallat小波边缘检测进行人脸的精确定位,利用矩值确定人脸的伸缩因子和旋转角度后进行归一化,有效的减少了光照不均匀和人脸旋转的影响。利用Matlab平台进行计算,实验结果表明,算法速度快,具有很高的精度和很强的鲁棒性。
This paper presents an efficient algorithm designed for detecting eyes with different poses in complex background to solve the problem of Nolllinear Illumination and different poses. In this algorithm, the Nonlinear Illumination effect is minimized by selecting suitable skin subspace and using a Mallat - Wavelet edge detection method, and the distortion is minimized by normalizing the human face based on the moment invariant vector. The result of experiment based on Matlab demonstrate the high efficiency of the algorithm in runtime and correct localization rate.
出处
《计算机仿真》
CSCD
北大核心
2009年第10期225-228,共4页
Computer Simulation
关键词
颜色空间
小波
矩
姿势
眼睛检测
Color space
Wavelet
Moment
Pose
Eye detection
