Abstract: An improved face recognition algorithm is proposed based on the combination of Modular 2DPCA and NSA. Non-parametric Subspace Analysis（NSA） transforms an image matrix to a vector which causes great dimensionality. NSA neglects the local feature of the image. The original images are divided into modular sub-images. NSA is utilized on the new pattern which is obtained by Modular 2DPCA to extract the final features from the sub-images. The new method considers the difference of between-classes and within-class while it extracts local feature of the image, and makes up for the flaw of the PCA. The experimental results obtained on the facial database ORL and XM2VTS show that the recognition performance of the new method is superior to that of the primary method of LDA and NSA.

Key words: Modular 2 Dimensional Principal Component Analysis（M2DPCA）, Non-parametric Subspace Analysis（NSA）, feature extraction, face recognition

摘要： 将非参数子空间分析方法（NSA）和模块化2DPCA方法相结合，提出了一种模块化2DPCA+NSA方法。NSA方法需将图像矩阵转化为向量后进行特征提取，导致数据维数很大，没有考虑到图像的局部特征，对图像矩阵进行分块，采用2DPCA进行特征提取，得到替代原始图像的低维新模式，施行NSA。该法能有效提取图像的局部特征，而由于考虑到类内、类间的差异，可弥补PCA 的缺陷。在ORL人脸库和XM2VTS人脸库上对LDA方法、NSA方法以及该方法分别进行了评价和测试，结果显示，所提方法在识别效果上优于LDA方法和NSA方法。

关键词: 模块化二维主元成分分析法（M2DPCA）, 非参数子空间分析方法（NSA）, 特征提取, 人脸识别