摘要
在分块2DPCA(Modular 2DPCA)算法的基础上,提出一种基于图像子块熵值加权的Modular 2DPCA算法(Entropy Modular 2DPCA)。Modular 2DPCA法直接计算测试图像与训练图像特征矩阵的距离,而Entropy Modular 2DPCA根据测试样本自适应确定图像子块的权值,增强包含分类信息多的子块权值,加入测试样本的信息,解决2DPCA人脸识别算法完全依赖人脸库的问题。将Entropy Modular 2DPCA算法、2DPCA算法以及Modular 2DPCA算法在ORL、自建人脸数据库上进行对比测试实验,实验结果表明,Entropy Modular 2DPCA算法具有良好的识别性能和计算速度,提高了对人脸姿态、光线、遮挡等问题的鲁棒性。
A weighted modular two-dimensional PCA algorithm based on image entropy（Entropy Modular 2DPCA）was proposed for face recognition based on modular two-dimensional PCA（Modular 2DPCA）.Different from those traditional modular two-dimensional PCA algorithms which directly calculate the distance of eigen matrix between test sample and training sample,EM2 DPCA self-adaptively determined sub-image weight according to test sample,so the weight including more information for classification was enhanced,and the problem that 2DPCA algorithm for face recognition depended on face databases was solved.EM2 DPCA,2DPCA and M2 DPCA were compared based on ORL and self-constructed face datasets.Results show that EM2 DPCA has good recognition performance and high calculation speed,and it enhances robustness to face gesture,illumination changing,shelter,etc.
出处
《计算机工程与设计》
北大核心
2016年第4期954-958,共5页
Computer Engineering and Design
基金
山西省自然科学基金项目(2013011017-4)
关键词
人脸识别
2DPCA
熵值
加权
分块
face recognition
two-dimensional PCA
entropy
weighted
modular
