Infrared face recognition based on radiative energy and sub-block DCT in wavelet domain

Abstract

A novel infrared face recognition method is proposed in this paper. In this method, according to Stefan-Boltzmann's law, the raw thermal images are transformed to the radiative energy images, and they are decomposed using two scales' discrete wavelet transform. The components of low frequency sub-bands are partitioned into sub-blocks, then, they are transformed by DCT. Each face is represented by the coefficients extracted from the sub-blocks in DCT domain. Finally, according to the discriminative power, different sub-blocks in DCT domain can be assigned different weights. Experiments demonstrate the method proposed perform very well on recognition rates as compared to other methods.