低分辨图像眼睛精确定位方法

Precise eye location in low resolution image

针对低分辨率图像中的人眼定位问题,提出一种融合主动表观模型和在线肤色纹理特征的低分辨图像中精确定位眼睛的方法。该方法通过在线人脸的肤色模板,结合眼睛局部区域的纹理特征,同时引入眼睛拟合模型,对上下眼睑边缘和眼角点进行精确拟合定位。考虑到虹膜区域的低分辨特点,使用圆形区域积分计算虹膜的精确位置,实现对虹膜中心点的精确定位。实验结果表明,该算法在低分辨率图像特别是网络摄像头采集的视频人脸图像中能够鲁棒实现眼睛区域的精确定位。

This paper presented a combined skin texture features model with active appearance model algorithm to realize the precise positioning of the low resolution face image. First of all, it considered the low resolution characteristics of the eye region and used an eye fitting model to extract the positioning of the eyelids. Then, it derived the online facial skin model to accurately locate the eye comers. Finally, it solved the iris detection based on an improved circle integral derivative. The experimental resuhs show that the proposed method not only obtains highly accurate location, but also has a faster calculating speed.