基于分块的2DPCA人脸识别方法

摘要

文章将分块理论与2DPCA方法相结合，研究分块二维主成分分析法（M-2DPCA）在人脸识别中的应用。对人脸图像矩阵进行分块，用形成的子图像矩阵直接构造总体散布矩阵并求解对应的特征向量，利用提取的特征向量对图像进行特征的提取与分析，进行人脸识别。基于Yale人脸数据库的实验显示，在相同训练样本和特征向量条件下，M-2DPCA比2DPCA算法具有更高的识别率。结论 M-2DPCA充分利用了图像的协方差信息，在人脸识别方面具有较高的识别率和鲁棒性方面，对进一步研究人脸识别具有重要的意义。%Aim: The block theory and two-dimensional principal component analysis (2DPCA) were combined, and the modular two-dimensional principal component analysis (M-2DPCA) was studied in face recognition. Methods: The original image matrix was divided into modular image matrixes , and the image covariance matrix was formed directly by using sub-image matrixes , and its eigenvectors were derived. The eigenvectors were used to extract and analyze image feature for face recognition. Results:The Experiments based on the Yale face database showed that it had a higher recognition rate of M-2DPCA than 2DPCA, under the same training specimens and eigenvectors. Conclusion: The information of image covariance matrix was fully utilized in M-2DPCA method , which had an admirable recognition rate and robustness on face recognition, and it was important to further research on face recognition.