基于选择性阈值取反和径向对称的瞳孔中心定位  

Pupil center location based on radial symmetry combined with selective threshold

作　　者：赵彦涛[1] 聂向荣[1] 罗军[1] 郑龙江[1] 张旭光[1] 李小俚[1] 

机构地区：[1]燕山大学电气工程学院,河北秦皇岛066004

出　　处：《光电子．激光》2016年第11期1208-1213,共6页Journal of Optoelectronics·Laser

基　　金：国家自然科学基金(61271409);河北省自然科学基金项目(C2014203212);河北省博士后科研项目择优(B2014003011)资助项目

摘　　要：提出一种选择性阈值取反与径向对称变换结合的算法,用于瞳孔中心的定位。提取人眼区域图像并转化为灰度图像,采用高斯平滑减小噪声的影响,对灰度图像进行阈值取反操作处理,去除眼睛周围像素的梯度变化对定位的影响。选择性阈值取反可综合利用直接取反和阈值取反的优点,进行径向对称变换投票得到瞳孔中心,实现瞳孔中心的定位。在BioID数据库及CASIAIrisV4进行实验验证的结果表明,本文提出的选择性阈值取反操作后进行径向对称变换的瞳孔中心定位可取得更高的精度,满足实时性的要求,可应用于人机交互的视线追踪系统。The pupil center location plays an important role in the human-compu ter interaction gaze tracking system.An algorithm that combines radial symmetry with image selective threshold reversion is presented to locate the pupil center.Firstly,the eyes on a face image are extracted by the relative position of eyes and the center of the face area.The color eye image should be converted into a gray image,then the image is smoothed using the Gaussian smoothing to reduce the influence of noise.The gray scale image is reversed with a threshold,which can remove the influence of pixels′s gradient changes of the eye surroundings.After th image selective threshold reversion,we can use the radial symmetry to vote t he pupil center.In the experiments,we use the BioID and CASIA-IrisV4datasets t o test the approach,which demonstrates that the proposal can achieve higher accuracy in pupil center localization,the method meets the real-time requirement,so w e can apply it to the human-computer interaction gaze tracking system.Abstract ：The pupil center location plays an important role m the human-computer mL-l-tc-Lu,, S,-- ing system. An algorithm that combines radial symmetry with image selective threshold reversion is presented to locate the pupil center. Firstly, the eyes on a face image are extracted by the relative position of eyes and the center of the face area. The color eye image should be converted into a gray image,then the image is smoothed using the Gaussian smoothing to reduce the influence of noise. The gray scale im- age is reversed with a threshold,which can remove the influence of pixelsrs gradient changes of the eye surroundings. A-ter th image selective threshold reversion, we can use the radial symmetry to vote the pupil center. In the experiments,we use the BioID and CASIA-IrisV4 datasets to test the approach, which demonstrates that the proposal can achieve higher accuracy in pupil center localization,the method meets the real-time requirement, so we can apply it to the human-computer interaction gaze tracki

关 键 词：高斯平滑 阈值取反 径向对称变换 投票 瞳孔中心 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]