自适应平滑自商图与多层分块的人眼定位算法

Eye location algorithm based on adaptive smoothing self-quotient image and hierarchical block

针对光照对眼睛定位产生的影响,提出了一种基于自适应平滑自商图像与多层分块的人眼定位算法。首先进行光照评估,利用自适应平滑自商图像(ASSQI)对光照条件差的进行处理,消除人脸图像的光照影响;然后运用眼睛区域梯度复杂度大的特征并结合积分图分块处理获得眼睛粗定位;最后对粗定位区域继续分块并利用质心得到准确的瞳孔中心。在JAFFE数据库、YaleB数据库和BioID数据库上测试表明该文算法有较高的定位正确率,对光照的鲁棒性较强。

An eye location algorithm based on adaptive smoothing self-quotient image and hierarchical block is proposed to locate eyes in this paper.This new algorithm involves three main steps.Firstly,illumination condition of the image is estimated,and Aadaptive Smoothing Self-Quotient Image（ASSQI） is introduced to correct illumination changes.Then,according to the characteristics of gray sharp changes,the candidate regions are located roughly by using integral image technique.Finally,accurate eyes location are determined by blocking and centroid.The experiment results in JAFFE,Yale B,and BioID face database show that this algorithm is effective and robust to illumination.