Skip to main content

Table 2 Verification rate at and in parenthesis the relative contribution that automatic registration has in comparison with manual registration on FRGC [4, Experiment ], comparing all registration methods using all face classifiers. The best automatic registration is achieved using user-independent SHR using low-resolutions, this often performs even better than manual registration.

From: Subspace-Based Holistic Registration for Low-Resolution Facial Images

Face Classifier FAR Viola-Jones (high resolution) MLLL + BILBO (high resolution) SHR (low-resolution) Manual
PCA Mah 1% 57.3% (8.9%) 67.4% (+1.3%) 68.1% (+2.0%) 66.2%
  0.1% 44.5% (5.8%) 52.9% (+2.6%) 54.0% (+3.3%) 50.3%
  0.01% 34.0% (3.4%) 40.9% (+3.4%) 42.2% (+4.7%) 37.5%
PCA MahCos 1% 73.2% (13.8%) 85.2% (1.7%) 87.9% (+2.0%) 87.0%
  0.1% 57.4% (10.8%) 68.2% (0.0%) 71.9% (+3.3%) 68.2%
  0.01% 39.7% (9.3%) 47.4% (+4.1%) 50.7% (+4.7%) 43.3%
Likelihood ratio 1% 86.7% (10.1%) 94.0% (2.8%) 95.9% (0.9%) 96.8%
  0.1% 76.9% (13.9%) 86.9% (4.7%) 91.0% (+0.2%) 90.8%
  0.01% 65.5% (14.8%) 77.2% (3.1%) 82.5% (+2.2%) 80.3%
Adaboost 1% 86.5% (8.4%) 93.5% (1.4%) 94.1% (0.8%) 95.0%
  0.1% 78.3% (10.5%) 87.1% (1.7%) 87.9% (1.0%) 88.9%
  0.01% 69.8% (11.1%) 78.9% (2.0%) 80.1% (0.8%) 80.9%