中文摘要：

为了减小环境温度对红外人脸图像的影响,本文提出了一种基于统计回归模型的红外人脸温度归一化方法.为了得到环境温度与红外人脸温谱图在对应像素点灰度之间的关系,将环境温度改变值和对应人脸上的温度变化值作为研究对象,利用统计回归方法对这两个对象进行二次多项式拟合即可得到环境温度变化和对应的人脸上温度变化的函数关系.通过得到的函数关系,建立归一化模型对红外图像进行温度归一化处理,减小环境温度对红外人脸识别的影响.实验结果表明：相对于归一化前的图像,温度归一化后的红外人脸图像与参考图像之间的信噪比有了明显改善,本文提出的归一化方法提高了红外人脸识别识别率.

英文摘要：

To reduce the impact of ambient temperatures on infrared face images,a normalization method of infrared images was proposed based on a statistical regression model.Firstly,we obtained the changes of both ambient temperatures and corresponding temperatures in face were obtained,which were used to establish a function through a second-order polynomial model.Then,the infrared images were normalized to reference ambient temperatures by the relevant function.Our experiments demonstrated that the images with temperature normalization have higher signal noise ratio than those without normalization,and that our method can significantly improve the infrared face-recognition performance.