Polar Topographic Derivatives for 3D Face Recognition: Application to Internet of Things Security

Abstract

We propose Polar Topographic Derivatives (PTD) to fuse the shape and texture information of a facial surface for 3D face recognition. Polar Average Absolute Deviations (PAADs) of the Gabor topography maps are extracted as features. High-order polar derivative patterns are obtained by encoding texture variations in a polar neighborhood. By using the and Bosphorus 3D face database, our method shows that it is robust to expression and pose variations comparing to existing state-of-the-art benchmark approaches.