Face recognition with discrete orthogonal moments

Abstract

In this chapter, an investigation has been conducted on the applicability of discrete orthogonal Hahn and Racah moments for face recognition regarding to robustness against lighting, facial expression, and face detail changes. The discrete Hahn and Racah moments are able to extract global facial features as well as local characteristics, thus providing the holistic, the component-based, and the fused approaches for feature representation. To classify the discrete orthogonal moment descriptors, the conventional nearest neighbour algorithm is employed with the Euclidean and Manhattan normalized distance metrics. The experimental results on the Olivetti Research Laboratory (ORL) database and University of Notre-Dame (UND) X1 biometric collection show that the recognition rates can reach 94% and 94.5% for the fused Hahn moments, and 92% and 94% for the combined global and local Racah moments, respectively.