通过分析在光照变化条件下现有的一些眼睛定位算法的不足,提出一种基于各向异性滤波的人眼定位方法。利用各向异性低通滤波器对输入图像进行滤波消除光照的影响,利用多结构形态学开闭运算突出眼睛区域特征,并采用相关系数法获得眼睛粗定位,最后使用眼睛精确算法校正眼睛的位置。实验结果表明,该算法对光照变化不敏感,定位准确率更高。
This paper first analyzed the shortages for some eye location algorithms to changes in illumination, proposed a method for human eye location based on the anisotropic filtering. To eliminate the effects of light, made use of the anisotropic low-pass filter to get the input image filtered, and used open and close operator of the multi-structure morphology to highlight the feature of the eye region. Then, got the rough position of the eye by the correlation coefficient method. Finally, corrected eye position by the accurate algorithm. The algorithm to changes in illumination is not sensitive and get higher positioning accuracy in experimental.