梯度估计在眼睛特征提取中的应用  

Gradient Estimation for Eye Feature Extraction

作　　者：楼京京[1] 

机构地区：[1]义乌工商学院计算机工程系,浙江义乌322000

出　　处：《计算机仿真》2008年第9期238-241,共4页Computer Simulation

摘　　要：眼睛特征的提取在表情分析、人脸合成和人脸识别等应用中起着非常重要的作用。目前广泛使用的眼睛特征提取算法是基于模板匹配的算法和基于特征点的算法。然而基于模板匹配的算法通常需要进行多个参数的选择,而且匹配过程非常耗时;特征点的算法则存在准确度较低的问题。针对以上问题提出了一种基于梯度信息提取虹膜中心、虹膜外轮廓和眼睛角点等眼睛特征的算法,算法用改进霍夫变换提取虹膜中心和外轮廓。在获得虹膜中心点后,参照中心位置选择眼睛角点候选区域并对其梯度进行分析。依据角点附近梯度分布与其它区域梯度分布不同的性质定位角点。仿真结果表明提出的算法能有效的提取眼睛特征,在测试数据上得到了比其它算法更高的准确率和更快的速度。Eye features extraction is a vital step for many applications such as facial expression analysis,face synthesis and face recognition.Currently,the widely used method is based on both template deformation or feature point.But template deformation based method needs to determine many parameters and the procedure of deformation is time-consuming,while the accuracy of feature point based method is relatively low.Focusing on the problems described above,an algorithm based on gradient information is presented to extract eye features including iris center,iris outer contour and eye corners from frontal images,.The proposed algorithm first extracts iris features using modified Hough Transform（HT）,then the gradient information of eye corner candidate area is analysed to locate the eye corner.The simulation results indicate that the proposed method can extract eye features effectively,and achieve better accuracy and faster speed than existing methods.

关 键 词：梯度估计 眼睛特征提取 霍夫变换 

分 类 号：TB851[一般工业技术—摄影技术]