Abstract: This paper proposes a novel face recognition algorithm based on Wavelet Packet Transform（WPT） and two-Dimensional Quaternion Principal Component Analysis（2D-QPCA）.Wavelet packet decomposition coefficients are combined to constitute the quaternion matrix，then 2D-QPCA is used to reduce dimensions and construct quaternion feature space.After that the space is divided into several sub-blocks，and each sub-block is classified according to the principle of the nearest neighbor.These sub-block classification results are voted to complete the ultimate face recognition.Compared with the traditional algorithm such as the PCA algorithm，experiment results on four face databases（Orl，Yale face database，etc）show that the proposed face recognition algorithm is superior and robust to the light and expression changes.

Key words: face recognition, quaternion, wavelet packet transform, two-dimensional principal component analysis

摘要： 提出了一种用小波包变换（WPT）和二维四元数主成分分析（2DQPCA）的灰度人脸图像识别算法。将对人脸灰度图像经小波包变换得到的分解系数构成四元数矩阵，通过2DQPCA实现数据降维并构造四元数特征空间，将其划分为若干子块，对每个子块根据最近邻算法进行分类并对分类结果投票，根据投票结果实现最终的人脸识别。该方法与PCA等传统方法在Orl、Yale等四个人脸数据库上的实验结果比较表明，该方法在识别率上有明显优势，且对光照、表情变化具有鲁棒性。

关键词: 人脸识别, 四元数, 小波包变换, 二维主成分分析