Pose-Invariant Face Recognition in Hyperspectral Images

Pose-invariant face recognition remains a challenging problem, especially when the pose change is large. Previous studies use either spatial or spectral information to address this problem. In this paper, we propose an algorithm that uses spatial and spectral information simultaneously to deal with large pose changes. We first learn 3D models from 2D images. We then use these 3D models to generate images in novel poses. Finally, we use spatial and spectral information to classify a test image. We demonstrate the effectiveness of the algorithm on a database of 200 subjects. Keywords—Face recognition, hyperspectral, Gabor filter, principal-component analysis (PCA), gradient descent