Combining local patch based descriptors with discriminant dictionary learning for improved face recognition

In this paper, we present a novel face recognition (FR) method that combines local patch based descriptors and dictionary learning underpinning the Fisher discriminant criterion. In our method, a face is first subdivided into several local regions and each local region is then represented using patch-based local descriptors. For a given local region, these extracted patch-based local descriptors are applied to discriminant dictionary learning for deriving sparse representation of each corresponding local region. The obtained sparse coefficient vectors from all local regions are then fused together to yield the so-called combined sparse coefficient vector. This can be achieved by using weighted feature fusion. Finally, the combined coefficient vectors are applied for dimensionality reduction technique. We incorporate our proposed algorithm into general FR pipeline and achieve encouraging results on CMU-PIE (92.09%) and XM2VTSDB (94.23%) datasets, compared to previously developed methods.