一种基于多条件综合验证的精确定位瞳孔方法  

An Accurate Pupil Positioning Method Based on Multi-condition Comprehensive Validation

作　　者：卢永峰[1] 屈志毅[1] 赵勇[1] 

机构地区：[1]兰州大学信息科学与工程学院,甘肃兰州730000

出　　处：《计算机科学与技术汇刊（中英文版）》2014年第4期127-131,共5页Transactions on Computer Science and Technology

摘　　要：本文首先使用梯度方向直方图（HOG）对瞳孔进行初次定位，提取感兴趣区域（ROI），缩小处理范围，去除部分噪声干扰，然后根据圆的几何特征检测瞳孔，对瞳孔进行再定位；最后依据瞳孔边界灰度梯度差最大的特征，逐个像素填充瞳孔边界，对已定位结果进行校正，在定位过程中，结合多种条件综合对瞳孔边界像素进行验证，保证边界像素的准确性，从而实现对不规则瞳孔边界的无误检测，达到精确定位瞳孔的目的。This paper firstly uses a gradient direction histogram （HOG） on the pupil for the initial positioning, extract the region of interest （ROI）, narrow processing range, remove part of the noise disturbance, and then based on the geometric feature of the circle to detect the pupil, re-position the pupil. Finally, according to feature of the difference of the maximum gradient of the pupil boundary, the pupil boundary is filled pixel by pixel. Make some corrections to the result of the positioning. In the positioning process, combine a variety of conditions to proof the pupil boundary pixel in order to ensure the accuracy of the boundary pixels, thereby achieve the accurate detection of the irregular pupil boundary, to the purpose of precise positioning of the pupil.

关 键 词：图像梯度方向直方图 ROI 几何特征 定位 

分 类 号：TP[自动化与计算机技术]