Computational Biometrics for Face Recognition

Joint work with: Aly Farag and Thomas Starr

The goal of this research was to propose a framework for face recognition at a distance based on texture and sparse-stereo reconstruction. At the Computer Vision and Image Processing (CVIP) Lab at the University of Louisville (UofL), we developed a biometric system capable of acquiring facial information at-a-distance to be integrated into surveillance/identity tasks in which active cooperation from the target may not be possible. The acquisition system consists of two CCD stereo cameras mounted on pan-tilt units with adjustable baseline.

The below figure shows the trilogy of the face recognition at a distance in CVIP Lab: image acquisition, leading to capturing objects in the field of view of the sensor; reconstruction, leading to mapping the captured objects into a form suitable for the final recognition step; which identifies the detected objects by correspondence with a dynamic database.

The proposed face recognition system framework.

The main focus was to develop an at a distance biometric modality based on facial recognition (FRAD). At the outset, most efforts were directed toward 2D facial recognition that utilizes the projection of the 3D human face onto the 2D image plane acquired by digital cameras. Challenges involving 2D face recognition are well documented in the literature. Intra-subject variations such as illumination, expression, pose, makeup, and aging can severely affect a face recognition system. To address pose and illumination, researchers have recently begun focusing on 3D face recognition. To that end, my colleagues and I at the University of Louisville developed a passive stereo acquisition setup to facilitate 3D facial surface reconstruction. We initially captured images from 30 subjects at various distances (3, 15, and 33 meters) indoors. Shape from sparse-stereo reconstruction was used to identify subjects, with acceptable results. We increased the number of subjects to 60 and captured images at 30m and 50m outdoors. In addition to sparse-stereo reconstruction, we performed dense 3D stereo reconstruction to aid in recognition, and used Active Appearance Model (AAM) for face alignment. Many existing facial recognition algorithms suffer in the presence of suboptimal lighting, pose, and many other factors. In our research, we overcame this problem through the use of constrained local models and texture-based features such as Gabor wavelets and local binary patterns.

Samples from our reconstruction results from the 3-, and 15-meter gallery and probe.

Related publications:

Mostafa Abdelrahman, Asem Ali, Shireen Y. Elhabian, Ham Rara, Aly A. Farag. A Passive Stereo System for 3D Human Face Reconstruction and Recognition at a Distance. IEEE CVPR Workshop on Biometrics, 2012, pp. 17-22.

Mostafa Abdelrahman, Shireen Y. Elhabian, Asem Ali and Aly A. Farag. Face Recognition at-a-Distance Using Texture, Sparse-Stereo, Dense-Stereo, Inteprnational Conference on Multimedia Technology (ICMT), pp. 6690-6695, 26-28 July 2011.

Ham Rara, Aly Farag, Shireen Y. Elhabian, Asem Ali, Mike Miller, Thomas Starr, Todd Davis. Face Recognition at-a-Distance Using Texture and Sparse-Stereo Reconstruction. Proc. of IEEE Fourth International Conference on Biometrics: Theory, Applications and Systems (BTAS), pp. 1221-1224, 2010.

Ham Rara, Asem Ali, Shireen Y. Elhabian, Thomas Starr, Aly A. Farag. Face Recognition at-a-Distance Using Texture, Dense- and Sparse-Stereo Reconstruction. Proceedings of the International Conference on Pattern Recognition (ICPR), pp. 1221-1224, 2010.

Ham Rara, Shireen Y. Elhabian, Asem Ali, Travis Gault, Mike Miller, Thomas Starr, Aly Farag. A Framework for Long Distance Face Recognition Using Dense- and Sparse-Stereo Reconstruction. 5th International Symposium on Visual Computing (ISVC09), Nov. 30 - Dec. 2, 2009, Las Vegas, Nevada, USA. , pp. 774-783.

Ham Rara, Shireen Y. Elhabian, Asem Ali, Mike Miller, Thomas Starr, Aly Farag. Face Recognition at a Distance Based on Sparse-Stereo Reconstruction. IEEE CVPR Biometrics Workshop, 2009, pp. 27-32.

Ham Rara, Shireen Y. Elhabian, Asem Ali, Mike Miller, Thomas Starr, Aly Farag. Distant Face Recognition Based on Sparse Stereo Reconstruction. IEEE International Conference on Image Processing (ICIP), Nov. 7 - Nov. 10, 2009, pp. 4141-4144.