In this paper an investigation of the validity of Gabor filter banks for feature extraction in a face recognition context is presented. Using the combined 2D/3D database collected by the Computer Vision Laboratory at the University of Notre Dame and the Colorado State University’s implementation of the Eigenfaces method, a comprehensive evaluation of the Log-Gabor filter bank representation is performed. Analysis of both the spatial and frequency domains is conducted to evaluate the distribution of discriminatory information.
展开▼