Hough变换在眼睛特征提取中的应用  

Hough transform for eye feature extraction

作　　者：黎云汉[1] 朱善安[1] 

机构地区：[1]浙江大学电气工程学院,浙江杭州310027

出　　处：《浙江大学学报（工学版）》2008年第7期1164-1168,共5页Journal of Zhejiang University：Engineering Science

摘　　要：为了准确地提取虹膜和眼睑轮廓等眼睛特征,提出了一种基于改进Hough变换的算法.该算法采用满足梯度要求的点对及其梯度信息确定输入图像中存在的圆,对于梯度反方向延长线相交且与交点距离相等的两点,定义它们属于同一个圆.在提取眼睑轮廓时,将候选点分为梯度方向向上和向下两部分并构造新的累加器分别确定上下眼睑所对应的圆,避免了上下眼睑的相互干扰.实验结果表明,算法在眼睛非闭合状态和正常光照下能准确地提取眼睛特征.与大多数Hough变换利用3个特征点来确定圆相比,算法通过引入梯度信息的方法,计算复杂度从O(N3)降低为O(N2).A robust algorithm based on modified Hough transform was proposed to accurately extract eye features including iris center and radius, point-pairs and their gradient information longing to the same circle if the reversing eyelid contours from frontal face images. The algorithm used to define circles in input image. Two points were regarded as beextension line of the two points＇ gradient orientation had an intersection and the distances between the intersection and the two points were equal. When extracting eyelid contours, the candidate points were divided into two parts based on gradient orientation and a new accumulation was constructed to define the circles corresponding to the upper and lower eyelid, thus the influence of each eyelid was avoided. Experimental results demonstrated that the algorithm accurately extracted eye features in normal illumination when eye was open. Compared with other algorithms using three points to define a circle, the algorithm reduces the combinatorial complexity from O（N^3） to O（N^2） using gradient information.

关 键 词：HOUGH变换 眼睛特征提取 图像处理 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]