基于DCT融合2DPCA与DLDA的人脸识别

传统的基于主成分分析的人脸识别需要将图像矩阵转化为向量,特征提取需要花费大最时间.二维主成分分析直接利用图像矩阵,特征提取速度快,但特征数量大,影响分类速度.因此,提出了一种基于离散余弦变换(DCT)的二维主成分分析(2DPCA)和直接线性判决分析(DLDA)结合的人脸识别方法.算法首先用DCT对人脸图像进行压缩并重建,然后利用2DPCA和DLDA对人脸图像进行特征提取.最后选用最近邻分类器进行分类.在ORL人脸库上的测试结果表明,与DLDA或2DPCA算法相比,算法具有更高的识别率.

Face Recognition Using 2DPCA and DLDA Based on DCT

In traditional principal component analysis(PCA) based face recognition technique,2D face image matrices must be transformed into 1D image vectors,much time was required to determine the eigenvectors.Two-dimensional PCA computed the features by using original image matrixes,but there were more features to represent an image in 2DPCA than PCA,which slowed down the classification speed.So,a new face recognition method was proposed,which used two-dimensional 2DPCA and direct-linear discriminant analysis(DLDA) ...