Face recognition from near-infrared images with convolutional neural network

Abstract

Owing to the vigorous development of face recognition, near-infrared (NIR) face recognition technology with light insensitivity has attracted increasing attention. However, the traditional methods for NIR face recognition feature the hand-crafted feature design. In this paper, we present a convolutional neural network (CNN) for NIR face recognition. CNN is a multiplayer feed-forward neural network which can automatically learn the features from the raw images and provide partial invariance to illumination, scale and deformation. Experimental results on PolyU-NIRFD database show that our proposed CNN architecture has higher recognition rate compared with the traditional recognition methods, such as Gabor-directional binary code (GDBC), Zernike moments and Hermite kernels (ZMHK).