Anmerkungen:
Diese Datenquelle enthält auch Bestandsnachweise, die nicht zu einem Volltext führen.
Beschreibung:
The proposed face alignment algorithm uses local gradient features as the appearance representation. These features are obtained by pixel value comparison, which provide robustness against changes in illumination, as well as partial occlusion and local deformation due to the locality. The adopted features are modeled in three discriminative methods, which correspond to different alignment cost functions. The discriminative appearance modeling alleviate the generalization problem to some extent.