摘要： 提出一种分块类增广PCA方法并应用于人脸识别中。对原始图像矩阵进行分块，对各个分块子图像施行自适应的CAPCA方法进行特征提取，将提取到的特征依次整合，从而达到降维的目的。该方法不仅能有效提取图像的局部特征，而且能适应不同的光照条件。实验结果表明，该方法在识别性能上优于CAPCA方法和分块PCA方法。

关键词: 人脸识别, 特征提取, 分块, 类增广PCA

Abstract: This paper presents a method called Sub-pattern Class-Augmented Principal Component Analysis(SpCAPCA) and its application in face recognition. The original images are divided into several sub-patterns by the presented approach. The SpCAPCA method is applied to the sub-patterns obtained from the previous step. The obtained features are combined according to a certain order. The dimension of the original images can be reduced. This approach can not only extract the local features of the images effectively, but also adapt the complicated illumination conditions. Experimental results show that the proposed SpCAPCA algorithm achieves better performance than other feature extraction methods, such as CAPCA, SpPCA.

Key words: face recognition, feature extraction, sub-pattern, Class-Augmented Principal Component Analysis(CAPCA)

中图分类号: 

• TP391.41