Abstract: In order to improve the performances of face recognition under non-uniform illumination conditions，a face recognition method combining illumination compensation with features extraction robust to change in lighting is proposed.Firstly，the original images are pre-processed by illumination compensation using Discrete Cosine Transform（DCT） in the logarithm domain.Secondly，cubic spline dyadic wavelet is applied to extract an approximiation subband and three detail subbands which are robust to illumination variations.And then，2DLDA is used to reduce dimension and four classifiers are constructed.At last，the result can be obtained by fusion of four classifiers.The proposed algorithm is tested on CAS-PEAL and YaleB face databases，and achieves 83.91% and 100% recognition rate respectively.The experiments prove that the proposed method is robust to varing lighting conditions.

摘要： 为进一步提高各种光照条件下的人脸识别精度，提出了一种将光照补偿和光照不变特征提取相结合的人脸识别方法。算法先应用对数域DCT进行光照补偿；然后，用三次样条二进小波分解提取一个低频子图和三个对光照变化鲁棒的边缘细节子图；接着，用二维线性判别分析进行特征降维并构造四个分量分类器；最后，通过多分类器融合规则进行融合分类。该文算法在CAS-PEAL人脸库光照子集上的实验达到了83.91%的识别率，在YaleB人脸库上则实现了100%的识别率，实验结果证明了该文算法对光照变换具有较好的鲁棒性。