Abstract: With illumination varying condition，face feature gotten from image is distorted nonlinearly by variant lighting intensity and direction，so face recognition becomes very difficult.According to the “common assumption” that illumination varies slowly and the face intrinsic feature（including 3D surface and reflectance） varies rapidly in local area，high frequency feature represents the face intrinsic structure.FABEMD is a fast and adaptive method of BEMD，and not using time-consuming plane interpolation computation.It can decompose the image into multi-layer high frequency images representing detail feature and low frequency images representing analogy feature.But a quantitative analysis that how much detail feature can be used for eliminating illumination variation can’t be made.So two measurements are proposed to quantify the detail feature，and with these measurement weights，FABEMD based multi-layer detail images matching can be done for face recognition under vary illumination.With PCA，the experiment results based on Yale face database B and CMU PIE face database show the method can get remarkable performance.

Key words: illumination invariant, Fast Adaptive Bidimensional Empirical Mode Decomposition（FABEMD）, Bidimensional Empirical Mode Decomposition（BEMD）, face recognition

摘要： 在光照变化的环境下，人脸识别因受到光照强度和方向的非线性干扰而变得困难重重。在人脸局部区域，光照的变化比较缓慢，而皮肤对光照的反射率特征变化比较快，可以认为光照变化是低频信号，而人脸本质特征是高频信号。FABEMD是一种快速自适应的BEMD（Bidimensional Empirical Mode Decomposition，二维经验模式分解）方法，它能够将图像分解为不同尺度的高频图像和低频图像，高频图像代表了人脸皮肤细节纹理特征，而低频图像则代表了轮廓特征。但是并不能定量判别什么样的高频信号以及多少高频信号能够用来消除光照影响，所以提出了两种衡量高频细节信息量的方法，将这些信息量的相对值来推算融合不同尺度的高频信号权重系数。基于Yale B人脸数据库的实验数据证明了所提方法能够取得很好的识别效果。

关键词: 光照不变性, 快速自适应的二维经验模式分解（FABEMD）, 二维经验模式分解（BEMD）, 人脸识别