基于子模式双向二维线性判别分析的人脸识别

针对表情和光照变化等对人脸识别影响的问题,提出一种基于子模式双向二维线性判别分析(Sub-pattern two-directional two-dimensional linear discriminant analysis,Sp-(2D)2 LDA)的人脸识别方法。该方法首先对原图像进行分块处理,并保持子块间的空间关系,然后对各个子训练样本集从行方向和列方向同时利用2DLDA进行特征抽取,最后把各个子特征矩阵拼接成一对应原始图像的特征矩阵,并采用最近邻分类器进行分类识别。在ORL及Yale人脸库上的试验结果表明,Sp-(2D)2 LDA有效降低了鉴别特征的维数,减少了表情和光照变化的影响,获得了较好的识别性能。

Face recognition based on sub-pattern two-directional two-dimensional linear discriminant analysis

To reduce the impacts of the variations of expression and illumination, a novel face recognition method based on sub-pattern two-directional two-dimensional linear discriminant analysis(Sp-(2D)²LDA) is presented in this paper. Firstly, Sp-(2D)²LDA divides the original images into smaller sub-images and keeps the spatial relationship between the sub-images. Secondly, it simultaneously applies 2DLDA to the subsets of the training samples in the row and column directions to extract local sub-features. Finally, the sub-features are synthesized into global features and nearest neighbor classifier is used for classification. The experimental results on Yale and ORL face databases show that the proposed Sp-(2D)²LDA method effectively reduce not only the dimension of the eigenvectors, but also the influence of variations in illumination and facial expression. Thus, the proposed method has better classification performances than the other related methods.