摘要
基于人脸图像的曲线奇异性及高维图像数据带来的计算复杂性,提出一种结合Curvelet变换与LPP的人脸识别方法。首先通过Curvelet变换对人脸图像降维,利用LPP将图像投影到最优子空间中,利用支持向量机进行分类识别,实验结果表明该算法的识别效果优于小波变换结合LPP方法、LPP方法。
Based on curves singularity of face image and the computational complexity caused by high-dimensional image data, proposes a new face recognition algorithm based on Curvelet transform and LPP. Applies curvelet transform dimensionality reduction for face image, uses LPP to project the image to the optimal subspace, applies support vector machine （SVM） for classification. Experimental results on ORL and Yale indicate that the performance of proposed method is superior to other methods, such as Wavelet transform combined with LPP and LPP method.
出处
《现代计算机》
2013年第23期30-33,共4页
Modern Computer
关键词
人脸识别
CURVELET
LPP
支持向量机
Face Recognition
Curvelet
LPP（Local Preserving Projection）
Support Vector Machine
