非均匀光照下人脸眼睛的定位方法

A Novel Method of Detecting and Locating Human Eyes under Nonlinear Illuminations

考虑非均匀光照下人脸眼睛的检测问题 ,提出了一种利用高频信息模板匹配方法从复杂图像中定位人脸眼睛的方法。选取 80幅光线较好的人脸图像 ,对它们做光线规范化 ,提取Gabor高频特征 ,构造模板。利用统计模式识别的原理 ,在眼睛的大致区域进行模板匹配 ,突出眉毛与眼睛这一整体的大致位置 ,然后进行二值投影 ,最终确定人眼的准确位置。大量实验表明 ,该算法具有很高的精度和很强的鲁棒性。

This paper focuses on detecting eye states under nonlinear illuminations, with a n automatic and real time method proposed. The method is based on high-freque ncy information pattern recognition and image projection. Firstly, 80 facial cha racter istic pictures will be got after matching the histogram and extracting the high- frequency information by Gabor wavelet. Secondly, the templates, made based on t he idea which is established by the priori knowledge, include the eyebrows and e yes. Thirdly, after matching the picture, a block including eyebrows and eyes wi ll be got. Finally, the eyes can be accurately located based on projection on th e block. Experimental results demonstrate the high efficiency of the algorithm i n runtime and correct localization rate.