Eyes localization is an important part of human face biometrics. A fast and effective automatic position approach of the human eyes is introduced. It detects face blocks using the AdaBoost cascade classifier based on Haar-like features,and marks out the rough areas of left and right eye-eyebrow regions.The Improved Projection Function(IPF) is applied to separate the eyebrows from the eye regions. The two eye regions are devided into a number of small image blocks respectively.The accurately positions of eyes are defined after using a united method based on the mean, complexity and the structure central measurement. This method gets a good performance about 99.1％ of correct positioning in the Normal sub-library of the CAS-PEAL face database. The experimental results show that it's a fast and effective eyes localization method.%眼睛定位是人脸生物特征识别的重要组成部分.提出了一种快速、高效的人眼定位方法,该方法先用基于Haar-like特征的AdaBoost联分类器检测出人脸块,粗定位左、右眉眼区域,采用改进的灰度投影法分别将两块区域中的眉毛部分分离掉,最后分别将左右眼区域分割成若干小图像块,联合图像块灰度均值、复杂度和结构居中度等信息精确定位人眼.该方法在中科院人脸库CAS-PEAL的Normal子库上表现优异,取得了约99%的定位正确率.实验结果表明,该方法速度快,定位效果好.
展开▼
