Fractional Multiset Coherent Super-Resolution Representation for Low Resolution Face Recognition

Abstract

In this paper, we address the problem of multiple resolution simultaneous learning in the limited training samples or noise disturbance cases and propose a novel fractional multiset partial least squares (FMPLS) approach for simultaneously dealing with multiset high dimensional data. The proposed FMPLS reconstructs the sample covariance matrices by fractional order spectral decomposition. Through using this FMPLS as a tool, we further present a new fractional multiset coherent super-resolution representation (FMCSR) method for low-resolution face recognition. Experimental results on two benchmark face databases demonstrate the effectiveness of the proposed FMCSR method.