计算机与现代化

• 图像处理 • 上一篇    下一篇

基于全局与局部特征融合的人脸识别方法

  

  1. 厦门大学信息科学与技术学院,福建厦门361005
  • 收稿日期:2014-01-16 出版日期:2014-03-24 发布日期:2014-03-31
  • 作者简介:兰佩(1989-),女,湖南益阳人,厦门大学信息科学与技术学院硕士研究生,研究方向:计算机视觉,模式识别,人脸识别; 方超(1986-),男,硕士研究生,研究方向:计算机视觉,图像处理,人脸识别。

Face Recognition Method Based on Global and Local Features Fusion

  1. School of Information Science and Technology, Xiamen University, Xiamen 361005, China
  • Received:2014-01-16 Online:2014-03-24 Published:2014-03-31

摘要: 针对现有预处理算法存在的缺陷及单一人脸特征在识别中的局限性,本文在基于双眼独立动态阈值的人脸预处理方法的基础上,研究全局特征PCA、2DPCA与局部特征LBP、Gabor,分析对比这几种特征的识别效果及适用情况;根据对这几种特征的研究分析,采用特征融合的方式对PCA和LBP特征进行融合;实验结果验证了在ORL库和ESSEX库上采用决策级融合的识别率优于特征级融合及单一特征的识别率。

关键词: 人脸识别, 全局特征, 局部特征, 2DPCA, LBP, Gabor, 特征融合

Abstract: In order to overcome the defects of current pre-processing method and the limitation of the sole facial feature recognition, four face recognition methods are studied based on the eyes-independent dynamic threshold-based algorithm of pre-processing .We extract global features using PCA and 2DPCA, and Gabor and LBP features for face local feature, analyze and compare the recognition rate of these features extraction methods and the situation which is suitable to these features. Based on the analysis of these features, feature fusion experiments are designed for PCA and LBP to improve the recognition accuracy. The experimental results show that the recognition rate is higher when using the decision fusion instead of the feature fusion and sole feature on the ORL and ESSEX database.

Key words: face recognition, global feature, local feature, 2DPCA, LBP, Gabor, feature fusion

中图分类号: