A BEMD based muti-layer face matching: From near infrared to visual images

Due to different light spectral, facial images captured in visible light are quite distinguished from that in near infrared, degrading the performance of the general methods of face recognition. In this paper, a new method based on BEMD to match facial images from near infrared to visible light is proposed. Since BEMD can decompose an image into multiple hierarchical components known as bidimensional intrinsic mode functions, recognition or verification from near infrared to visual images can be carried out on muti-layer rather than simply one-to-one match. Besides, intrinsic structure invariant to spectral changes contained in images between them can be extracted using BEMD which greatly enhances the recognition or verification rate. Experiments show promising result of our methods.