An Improved SDA Approach with DFB Preprocesing for Face Recognition

Over the last decade, major advances have occurred in face recognition and a large number of algorithms have been proposed to achieve acceptable recognition rates under controlled conditions. However, a successful application under real world conditions remains a challenge. To address this problem, a new method for face recognition is presented and is based on Subclass Discriminant Analysis with a Directional Filter Bank pre-processing. Experimental results show that the DFB pre-processing step improves the original SDA method and attractive recognition rates are achieved.