基于眼睛梯度对特征的人脸检测方法

Face Detecting Based on Grads Couple Feature of Eyes

该文提出了一种利用眼睛灰度变化梯度对特征来检测人脸的方法。在分析人脸图像中发现,眼球的灰度级较低,而眼球周围白色区域的灰度级较高,两个区域在其交界处灰度级产生强烈突变。该文利用图像灰度变化梯度值构造梯度和特性值,通过寻找相等的梯度和特性值来实现从图像中找出对应眼睛的位置,定位可能的人脸区域,再计算其对称性后确定人脸特征的存在,更进一步验证检测的人脸区域。实验证明,此方法能迅速准确地从复杂背景图像中检测出人脸,而且对多人脸图像同样有效。

This paper presents a method of face detection based on grads couple feature of eyes.From analysing face image,it is easy to find that the gray level is lower in eyeball than the gray level of in white region around the eyeball,The gray level varies dramatically on the boundary between them.In order to get the feather,the first step is to create the image with grads;then chain the grads to description that is called grads sum.The next step is to find possible eyes in an image utilizing grads couple feature of eyes and locates possible face region.After that, their symmetries can be calculated and the possible face can be verified from the possible existing facial features.At last, the experimental results prove that this method can detect face in an image with complex background rapidly and accurately.Furthermore,it can perform well for multiple faces.