• Laser & Optoelectronics Progress
  • Vol. 55, Issue 3, 031009 (2018)
Yueping Kong1,*, Xia Liu1, Xinqian Xie1, and Fengjie Li1
Author Affiliations
  • 1 College of Information and Control Engineering, Xi'an University of Architecture and Technology, Xi'an, Shaanxi 710055, China
  • 1 Tianjin Electronic Information College, Tianjin 300350, China
  • show less
    DOI: 10.3788/LOP55.031009 Cite this Article Set citation alerts
    Yueping Kong, Xia Liu, Xinqian Xie, Fengjie Li. Face Liveness Detection Method Based on Histogram of Oriented Gradient[J]. Laser & Optoelectronics Progress, 2018, 55(3): 031009 Copy Citation Text show less
    Imaging process of real face and photo face
    Fig. 1. Imaging process of real face and photo face
    Comparison of real face, photo face, and their HOG. (a) real face, (b) gradient image of (a), (c) HOG of (b), (d) photo face, (e) gradient image of (d), and (f) HOG of (e)
    Fig. 2. Comparison of real face, photo face, and their HOG. (a) real face, (b) gradient image of (a), (c) HOG of (b), (d) photo face, (e) gradient image of (d), and (f) HOG of (e)
    Schematic of HOG feature extraction
    Fig. 3. Schematic of HOG feature extraction
    Block diagram of HOG feature extraction
    Fig. 4. Block diagram of HOG feature extraction
    Reference feature distributions of average features of (a) real face and (b) photo face
    Fig. 5. Reference feature distributions of average features of (a) real face and (b) photo face
    Similarity distributions between samples and reference features. Similarity distributions of 1000 samples to (a) HT-mean and (b) HF-mean
    Fig. 6. Similarity distributions between samples and reference features. Similarity distributions of 1000 samples to (a) HT-mean and (b) HF-mean
    Block diagram of human face liveness detection
    Fig. 7. Block diagram of human face liveness detection
    Example of error samples
    Fig. 8. Example of error samples
    ExtractionmethodFeaturedimensionTP /%TN /%Detectionaccuracy /%
    Proposed5095.8098.2097.00
    HOG4892.6094.2093.40
    Table 1. Test results of feature selection
    ExtractionmethodFeaturedimensionTP /%TN /%Detectionaccuracy /%
    Geometricalcharacteristic[24]13192.1285.5888.85
    LBP[5]5993.5594.0693.87
    Fourierspectrum+LBP[1]91100.0092.3396.16
    GLCM+waveletcharacter[19]1297.0396.8896.97
    Proposed5095.8098.2097.00
    Table 2. Recognition rate of different feature extraction methods