河北大学学报(自然科学版) ›› 2016, Vol. 36 ›› Issue (2): 204-209.DOI: 10.3969/j.issn.1000-1565.2016.02.016

• • 上一篇    下一篇

基于变尺度DCT和PCA的人脸识别方法

张欣,刘海,宋卓展   

  • 收稿日期:2015-10-19 出版日期:2016-03-25 发布日期:2016-03-25
  • 作者简介:张欣(1966—),男,河北承德人,河北大学教授,博士,主要从事机器视觉、图像处理、模式识别和故障诊断方向研究. E-mail:zhangxin@hbu.edu.cn
  • 基金资助:
    国家自然科学基金资助项目(61203160)

Variable scale face recognition based on DCT and PCA

ZHANG Xin,LIU Hai,SONG Zhuozhan   

  1. Electronic Information Engineering College, Hebei University, Baoding 071002, China
  • Received:2015-10-19 Online:2016-03-25 Published:2016-03-25

摘要: 针对实际应用中要对不同尺度的人脸图像进行匹配和识别的需求,提出一种基于变尺度DCT(离散余弦变换)结合PCA(主成分分析法)的人脸识别方法.首先将不同尺度的人脸图像进行DCT变换,然后提取出相同的DCT系数,再将提出的DCT系数进行PCA降维,最后进行匹配识别.通过对ORL人脸库及实验采集的多尺度人脸图像进行实验,证明了所提方法对不同尺度图像的匹配识别有效,并测算了DCT变换后的修剪尺度及PCA特征维数对该方法识别准确率的影响.

关键词: 人脸识别, 变尺度, DCT, PCA, 匹配

Abstract: A variable scale face images recognition method is proposed, based on the DCT binding the PCA. First, different scale face images are transformed by DCT. Then, the same DCT coefficients are extracted;and the dimension of the DCT coefficients is reduced by PCA. Finally, matching and recognition of images is done. By the ORL and collected face images, it is proved that the method of recognition variable scale images is efficient. Effective values of DCT transformed coefficients and PCA feature dimension are measured.

Key words: face recognition, variable scales, discrete cosine transform, principal component analysis, match

中图分类号: