Enhancing face recognition using Directional Filter Banks

Abstract

Face recognition is an increasingly important problem in biometric applications; consequently many recognition algorithms have been proposed during the last three decades. It is accepted that the use of a pre-processing step can extract more discriminating features and increase the classification rates. Although, Gabor filters have been widely employed they do not provide satisfying classification results. This paper proposes the use of directional filters as a pre-processing step to demonstrate that a Directional Filter Bank is capable of enhancing existing face recognition classifiers such as PCA, ICA, LDA and SDA. The proposed method is tested using two different databases: the Yale face database and the FERET database. Experimental results demonstrate that the pre-processing phase enhances the classification rates. A comparative study has also been carried out to demonstrate that a DFB based classification outperforms a Gabor type one.