Expression-insensitive three-dimensional face recognition using holistic and local features

This paper propose a fully automatic face recognition method and demonstrate its performance on the FRGC v2 data. A 3D facial representation based on the Gabor Transform was used for 3D face recognition in this paper. It is effective and efficient to represent the geometrical structures in facial image. The gabor feature is used to form a rejection classifier, which quickly eliminates a large number of candidate gallery faces. The remaining faces are then verified using forehead region Iterative Closet Point (ICP) matching method. This paper use holistic and local features into recognition and obtain good performance. This method achieves the rank one recognition rate of 97.99% on All vs. All experiment.