复杂背景下的多阈值人眼定位新算法*

提出了一种基于复杂背景下的多阈值人眼定位算法。该算法首先对图像进行光线补偿，减小由于光照不均等因素造成的影响；然后根据对人眼灰度特征的分析来确定3个人眼二值化阈值，从而分离出人眼候选区域；最后根据人脸的几何特征、人眼位置判定准则以及计算二维相关系数来排除非人眼区域，从而定位双眼。实验结果表明，该算法能在不同条件下准确定位人眼，且具有较快的定位速度和较高的检测率。

Eyes location based on multi-threshold in complex background

A novel algorithm for location of human eyes based on multi-threshold in the complex background is proposed. In this algorithm, the image is processed by light compensation so as to reduce the influence of unbalanced light firstly. Then through the analysis of the gray features of eyes, three binarization thresholds which are used to segment the alternative area of eyes can be determined. Finally, according to the geometric facial features and criteria for the eye position, eyes can be located by calculating two-dimensional correlation coefficient. The experimental results demonstrate that the proposed algorithm can detect the eyes in different conditions with high efficiency and rate of location.