Frontal Face Authentication Through Creaseness-Driven Gabor Jets

One of the most successful techniques for frontal face authentication is based on matching image graphs that represent faces. The local identity information is stored at node graph locations through vectors of multi-scale and multi-orientation Gabor image responses. In this work we present an efficient way to process f¡ontal faces by automatically locating graph nodes at selected crease pointsThis method does not rely on rigorous face alignment, increases speed in the pre-processing step and ¡esults in a face-graph representation that obtains comparable results to other methods reported in the literature over the XM2VTS DataBase.