人脸识别中光照补偿参数及必要性的判定方法*

针对实际人脸识别应用中大量人脸无须作光照补偿的情况,分别用2DPCA和Fisher两种人脸识别方法对FRVT标准人脸库进行识别,得出准确的影响人脸识别的光照方向参数,并提出了一种简单快速的判断是否需补偿的方法.实验证明,经过光照补偿判定后进行的人脸识别实验,需补偿人脸数目大量减少,运算量大大降低,人脸识别速度可以提高近四倍.

Compensated Illumination Direction Estimation Method in Face Recognition Based on 2DPCA and Fisher

To solve the problem that most of faces to recognize need not to be compensated in fact, this paper recognized faces in standard FRVT face database. Based on 2DPCA and Fisher methods respectively, it achieved the parameters on illumination influence to face recognition through the experiments. Meanwhile it proposed a simple and fast method to identify if the face under illumination should be compensated. The experimental results show that compensated faces and computation complexity are decreased greatly after compensation estimation. The recognition speed is up to four times than the traditional methods.