基于对称变换与高斯微分的人脸定位新方法

A New Face Localization Method Based on Symmetry Transform and Gaussian-Derivatives

在实际人脸识别系统中 ,复杂背景、无控制光照及成像质量对人脸的准确定位造成了严重影响 .本文针对上述情况提出了一种室内自然环境下人脸准确定位的新方法 .该方法充分利用了人脸具有的强对称性与三维特性 ,运用高斯微分求图像边缘 ,再对广义对称变换及径向对称变换加以规则限制 ,实现了人脸眉心的准确定位及对尺度因子的估计 ,进而实现了人眼的准确定位 ,具有很强的稳健性 .

In real face recognition systems,face localization is always heavily affected by complex background,uncontrolled lighting condition and low imaging quality.To solve these problems,a novel face localiztion method in natural indoor environment is presented.The main idea is to use the strong symmetry and the three dimensional property of human face.Firstly,edge map of original image is obtained by using the Gaussian derivative operator.Secondly,we introduce some novel rules to the generalized symmetry tranform and radial symmetry transform to localize the glabellum,and then to estimate the scale factor.Finally,accurate eyes localization is realized with similar idea.The results of some experiments are also given which show the robustness and effectiveness of this method.