A novel illumination compensation algorithm for face recognition under non-uniform illumination conditions was presented.Firstly,the face images were transformed from spatial domain to logarithm domain,and the 2-dimensional wavelet packet transform was calculated.Secondly,coefficients of some sub-band images were discarded for face illumination compensation in logarithm domain.The effect of face illumination compensation was denoted by face recognition rate because the intention of illumination compensation was to improve the performance of face recognition.The proposed approach was compared with the Discrete Cosine Transform (DCT) method in logarithm domain on Yale B face database.The experimental results show that the average error rate of the proposed approach can reach 0.53%,which is superior to the DCT method for face illumination compensation in logarithm domain.