Computer Engineering and Applications ›› 2020, Vol. 56 ›› Issue (6): 147-152.DOI: 10.3778/j.issn.1002-8331.1811-0364

Previous Articles     Next Articles

Research on HOG Feature Extraction Algorithm Weighted by Information Entropy

LIN Kezheng, ZHANG Yuanming, LI Haotian   

  1. School of Computer Science and Technology, Harbin University of Science and Technology, Harbin 150080, China
  • Online:2020-03-15 Published:2020-03-13



  1. 哈尔滨理工大学 计算机科学与技术学院,哈尔滨 150080


According to the different information entropy in different parts of face image, the influence of different factors on the recognition degree are different, this paper proposes an information entropy weighted HOG feature extraction method. The facial image to be identified is divided into blocks, HOG feature extraction on the block of the image, and then this paper calculates the information entropy of each image contained as weight coefficient to each block in the formation of new HOG features, the features are reduced by PCA algorithm, and the HOG features of information entropy weighting are obtained. The contrast experiment on ORL and YALE shows that this method not only has higher recognition accuracy than other traditional recognition methods, but also has good robustness and effectiveness for transforms of illumination, face pose and expression.

Key words: face recognition, feature extraction, information entropy, Histogram of Oriented Gradients(HOG), Principal Component Analysis(PCA)



关键词: 人脸识别, 特征提取, 信息熵, 梯度直方图(HOG), 主成分分析(PCA)