一种自适应加权变形的2DPCA人脸识别方法

Adaptive Weighted Variational 2DPCA for Face Recognition

为了充分利用图像矩阵的局部信息和更多的鉴别信息,以提高2DPCA的识别率,提出了一种自适应加权变形的2DPCA人脸识别方法。该方法将人脸图像矩阵分块,然后利用变形的2DPCA方法提取特征,接着自适应地计算每个分块在分类中的权值,最后根据类别的权值大小进行分类。在ORL人脸库中进行的实验研究表明,该方法在正确识别率和识别时间上更优于传统的2DPCA和模块化2DPCA。

For full using of local information and more discriminative information of the image matrix to improve the recognition rate of 2DPCA,an adaptive weighted variational 2DPCA for face recognition was presented,which partitions its subpatterns from an original whole image,extracts features from them in the variational 2DPCA,adaptively computes the weight of each subpattern in classification,finally makes classiffication in terms of weight of each classification.The experiments on ORL face bases show this algorithm is also superior to the traditional 2DPCA and Modular 2DPCA in terms of the recognition accuracy and recognition time.