在复杂的光照条件下人脸表面的光分布不规律，会出现局部过亮、过暗和反光等情况，会影响人眼定位的准确性。为了提高在复杂光照影响下人眼定位的准确率，本文提出一种先定位人脸再定位人眼的方案。首先采用Haar - like特征的Ada Boost级联分类器检测出人脸区域，对所得人脸部分进行自适应连续均值量化变换增强，然后结合连通域的分析对人眼进行定位。实验结果表明，该方法运算速度快，定位效果好。%In complex illumination, the uneven distribution of light on the face may lead to bright, dark or re-flection on part of the face image, which then subsequently brings about great impact on eye location. In order to improve the positioning accuracy of the human eye under complex illumination, the paper proposes a method which in the first place locates the face and then the eyes. The face can be detected based on Haar feature of Ada boosted cascade classifier. The face region is enhanced by the adaptive continuous mean quantization transform image enhancement algorithm. Then the eyes can be positioned by combining with the connected domain analysis. The results show that the proposed method has high detection speed and good results.
展开▼
