Periocular Face Recognition
 

We have studied the periocular face recognition from two aspects: (1) direct classifier learning and matching based upon the periocular region, (2) hallucinate the full face from the periocular region so that full face matchers and classifiers can be utilized.

 

Each one takes advantage of the merits given by the periocular region such as being the most discriminant part on human faces; most tolerant to age variations; most tolerant to expression variations; etc.

Hallucination
 
Hallucinating the Full Face from the Periocular Region via Dimensionally Weighted K-SVD
F. Juefei-Xu, D.K. Pal, and M. Savvides
Biometrics Workshop, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2014.
 
@INPROCEEDINGS{Felix_cvprw14_hallucinate,
author={F.~Juefei-Xu and Dipan~K.~Pal and M.~Savvides},
booktitle={Computer Vision and Pattern Recognition (CVPR) Workshops, 2014 IEEE Conference on},
title={{Hallucinating the Full Face from the Periocular Region via Dimensionally Weighted K-SVD}},
year={2014},
month={June},
pages={1-8},
}
 

In this work, we have developed a coupled dictionary learning approach called DW-KSVD for hallucinating the full face from the periocular region.

 

The proposed method differs from the existing methods in the sense that dimensions corresponding to the periocular region are weighted more during the optimization, such that the learned dictionary are more failthful towards the periocular region, and will enable better full face hallucination.

The dictionary learning for the full face and the periocular region are shown above respectively. The joint learning carries out by combing a weighted sum of the two aforementioned objectives. The parameter Beta controls how failthful the reconstruction is to the periocular region.
 
By stacking them together, the joint dictionary learning can be conducted effectively and efficiently using standard KSVD algorithm.
Spartans
 
Spartans: Single-sample Periocular-based Alignment-robust Recognition Technique Applied to Non-frontal Scenarios
F. Juefei-Xu, K. Luu, and M. Savvides
IEEE Transactions on Image Processing (TIP), vol. 24, no. 12, pp. 4780-4795, Dec. 2015.
 
@ARTICLE{Felix_tip15_spartans,
author={Juefei-Xu, F. and Luu, K. and Savvides, M.},
journal={IEEE Trans. on Image Processing},
title={{Spartans: Single-sample Periocular-based Alignment-
robust Recognition Technique Applied to Non-frontal
Scenarios}},
year={2015},
month={Dec},
volume={24},
number={12},
pages={4780-4795},
}
 

In this work, we have studied a single-sample, periocular-based alignment-robust face recognition technique that is pose-tolerant under unconstrained face matching settings. 

 

The highlight of this work is that, from a single-sample face image, we can generate new face images under a wide range of 3D rotations using 3DGEM. Then, we focus on the periocular region to extract overcomplete features. The classifier learning is through subject-dependent advanced correlation filters for pose-tolerant non-linear subspace modeling in kernel feature space.

© Carnegie Mellon University Biometrics Center 2015

  • s-twitter
  • s-facebook