Abstract: This paper proposes a new face recognition approach combining a Bayesian probabilistic model and Gabor filter responses.Since both the Bayesian algorithm and the Gabor features can reduce intrapersonal variation through different mechanisms，this paper integrates the two methods to take full advantage of both approaches.Firstly utilizing the convolution of the key points and the Gabor filters to extract features，and then 2DPCA is used to decrease the dimension of the intra-face and extra-face difference space in the Gabor feature space.Lastly use Bayesian method for face recognition.The experimental result shows that the method has the advantages of simple computation and high recognition rate under different expression and illumination by comparative experiment on AR and FERET face data.

Key words: face recognition, Gabor transform, Two-dimensional Principle Component Analysis（2DPCA）, Bayesian theory

摘要： 由于Gabor小波和贝叶斯方法都可以通过不同的机制来减少类内差异，提出了融合Gabor和贝叶斯的人脸识别方法。该方法首先通过人脸图像特征点与Gabor滤波器的卷积来提取特征，借鉴“作差法”形成“类内差”和“类间差”空间，并用2DPCA对差异空间进行降维，最后用贝叶斯方法进行分类。通过在AR和FERET人脸库上的实验表明，与传统的方法相比较，该方法降低了运算量，提高了识别率，对具有表情及光照变化的人脸具有较高的识别率。

关键词: 人脸识别, Gabor变换, 二维主元分析（2DPCA）, 贝叶斯方法