Abstract: This paper uses PCA and 2DPCA to do the face recognition，and obtains dimensions of eigenvalues and recognition accuracy by using different pre-set thresholds of biggest eigenvalues，and gives the optimalizing dimensions and the threshold of two above methods.Then，using the image rotation technology by bilinear interpolation algorithm to add the numbers of the samples when there is small sample size for same person.It gets higher recognition accuracy and solves the small sample size problem in some degree.If some new predicted information of the small sample images can be generated for same person such as increasing the different expressions and depth of the face for each sample，the effects of the test will be more obvious.

Key words: Principal Component Analysis（PCA）, two Dimensional Principal Component Analysis（2DPCA）, dimension, small samples size, bilinear interpolation, image rotating

摘要： 通过应用PCA及2DPCA算法进行人脸识别，得到了在取不同特征值门限情况下的特征提取维数和识别率，给出了以上两种算法最优特征提取向量的维数和最大特征值门限，并在此基础上应用双线性差值图像旋转处理技术，增加了同一个人较少训练样本情况下的训练样本数量，提高了识别率，从一定程度上解决了小样本问题。如果能从小样本图像中生成出一些新的预测信息，例如，增加同一个训练样本的不同的表情，或改变样本表情的深度，实验的效果可能更加明显。

关键词: 主成分分析, 二维主成分分析, 特征向量维数, 小样本, 双线性插值, 图像旋转