计算机工程与应用 ›› 2008, Vol. 44 ›› Issue (23): 197-200.DOI: 10.3778/j.issn.1002-8331.2008.23.059

• 图形、图像、模式识别 • 上一篇    下一篇

基于多姿态人脸图像合成的识别方法研究

孙志远1,吴小俊1,王士同1,杨静宇2   

  1. 1.江南大学 信息工程学院,江苏 无锡 214122
    2.南京理工大学 计算机科学与技术学院,南京 210094
  • 收稿日期:2007-10-18 修回日期:2008-01-16 出版日期:2008-08-11 发布日期:2008-08-11
  • 通讯作者: 孙志远

Face recognition based on synthesis of face images with pose variations

SUN Zhi-yuan1,WU Xiao-jun1,WANG Shi-tong1,YANG Jing-yu2   

  1. 1.School of Information Technology,Jiangnan University,Wuxi,Jiangsu 214122,China
    2.School of Computer Science and Technology,Nanjing University of Science and Technology,Nanjing 210094,China
  • Received:2007-10-18 Revised:2008-01-16 Online:2008-08-11 Published:2008-08-11
  • Contact: SUN Zhi-yuan

摘要: 为了解决多姿态人脸识别问题,提出基于独立成分分析(ICA)进行正面人脸合成的新方法。首先利用ICA和PCA提取不同姿态人脸的特征子空间,然后利用通过训练得到的姿态转换矩阵合成其相对应的正面人脸图像,实验表明ICA人脸识别算法要优于PCA人脸识别算法,并在此基础上用小波对人脸图像进行预处理,据姿态转换矩阵得到的正面人脸特征系数直接进行分类比较,识别率得到了很大的提高。

关键词: 独立成分分析, 小波变换, 姿态转换矩阵

Abstract: A new frontal face synthesis method based on Independent Component Analysis(ICA) is proposed to deal with the problem of face recognition with pose variations.First,the feature subspaces are formed from different pose images using ICA and PCA.Then the pose image is transformed into its corresponding frontal face image using the transformation matrix predetermined by learning.Based on this foundation,the wavelet transform is applied to face images,and the authors compare the frontal face feature coefficients directly which are obtained by using the transformation matrix with the feature coefficients of the original face images,the recognition rate is improved greatly.

Key words: Independent Component Analysis(ICA), wavelet transform, pose transformation