Pose Normalization via Learned 2D Warping for Fully Automatic Face Recognition

Arkshay Asthana, Micahel J. Jones, Tim K. Marks, Roland Goecke, Kinh H. Tieu

Research output: A Conference proceeding or a Chapter in BookConference contribution

23 Citations (Scopus)

Abstract

We present a novel approach to pose-invariant face recognition that handles continuous pose variations, is not database-specific, and achieves high accuracy without any manual intervention. Our method uses multidimensional Gaussian process regression to learn a nonlinear mapping function from the 2D shapes of faces at any non-frontal pose to the corresponding 2D frontal face shapes. We use this mapping to take an input image of a new face at an arbitrary pose and pose-normalize it, generating a synthetic frontal image of the face that is then used for recognition. Our fully automatic system for face recognition includes automatic methods for extracting 2D facial feature points and accurately estimating 3D head pose, and this information is used as input to the 2D pose-normalization algorithm. The current system can handle pose variation up to 45 degrees to the left or right (yaw angle) and up to 30 degrees up or down (pitch angle). The system demonstrates high accuracy in recognition experiments on the CMU-PIE, USF 3D, and Multi-PIE databases, showing excellent generalization across databases and convincingly outperforming other automatic methods.
Original languageEnglish
Title of host publicationProceedings of the British Machine Vision Conference (BMVC 2011)
EditorsJesse Hoey, Stephen McKenna, Emanuele Trucco
Place of PublicationDundee, United Kingdom
PublisherBMVA Press
Pages1-11
Number of pages11
ISBN (Print)978190172543X
DOIs
Publication statusPublished - 2011
Event22nd British Machine Vision Conference BMVC 2011 - Dundee, Dundee, United Kingdom
Duration: 29 Aug 20112 Sep 2011

Conference

Conference22nd British Machine Vision Conference BMVC 2011
CountryUnited Kingdom
CityDundee
Period29/08/112/09/11

Fingerprint

Face recognition
Experiments

Cite this

Asthana, A., Jones, M. J., Marks, T. K., Goecke, R., & Tieu, K. H. (2011). Pose Normalization via Learned 2D Warping for Fully Automatic Face Recognition. In J. Hoey, S. McKenna, & E. Trucco (Eds.), Proceedings of the British Machine Vision Conference (BMVC 2011) (pp. 1-11). Dundee, United Kingdom: BMVA Press. https://doi.org/10.5244/C.25.127
Asthana, Arkshay ; Jones, Micahel J. ; Marks, Tim K. ; Goecke, Roland ; Tieu, Kinh H. / Pose Normalization via Learned 2D Warping for Fully Automatic Face Recognition. Proceedings of the British Machine Vision Conference (BMVC 2011). editor / Jesse Hoey ; Stephen McKenna ; Emanuele Trucco. Dundee, United Kingdom : BMVA Press, 2011. pp. 1-11
@inproceedings{0377f43965014455a7407f2f8a971e77,
title = "Pose Normalization via Learned 2D Warping for Fully Automatic Face Recognition",
abstract = "We present a novel approach to pose-invariant face recognition that handles continuous pose variations, is not database-specific, and achieves high accuracy without any manual intervention. Our method uses multidimensional Gaussian process regression to learn a nonlinear mapping function from the 2D shapes of faces at any non-frontal pose to the corresponding 2D frontal face shapes. We use this mapping to take an input image of a new face at an arbitrary pose and pose-normalize it, generating a synthetic frontal image of the face that is then used for recognition. Our fully automatic system for face recognition includes automatic methods for extracting 2D facial feature points and accurately estimating 3D head pose, and this information is used as input to the 2D pose-normalization algorithm. The current system can handle pose variation up to 45 degrees to the left or right (yaw angle) and up to 30 degrees up or down (pitch angle). The system demonstrates high accuracy in recognition experiments on the CMU-PIE, USF 3D, and Multi-PIE databases, showing excellent generalization across databases and convincingly outperforming other automatic methods.",
keywords = "Pose normalization, Face recognition, 2D warping",
author = "Arkshay Asthana and Jones, {Micahel J.} and Marks, {Tim K.} and Roland Goecke and Tieu, {Kinh H.}",
year = "2011",
doi = "10.5244/C.25.127",
language = "English",
isbn = "978190172543X",
pages = "1--11",
editor = "Jesse Hoey and Stephen McKenna and Emanuele Trucco",
booktitle = "Proceedings of the British Machine Vision Conference (BMVC 2011)",
publisher = "BMVA Press",

}

Asthana, A, Jones, MJ, Marks, TK, Goecke, R & Tieu, KH 2011, Pose Normalization via Learned 2D Warping for Fully Automatic Face Recognition. in J Hoey, S McKenna & E Trucco (eds), Proceedings of the British Machine Vision Conference (BMVC 2011). BMVA Press, Dundee, United Kingdom, pp. 1-11, 22nd British Machine Vision Conference BMVC 2011, Dundee, United Kingdom, 29/08/11. https://doi.org/10.5244/C.25.127

Pose Normalization via Learned 2D Warping for Fully Automatic Face Recognition. / Asthana, Arkshay; Jones, Micahel J.; Marks, Tim K.; Goecke, Roland; Tieu, Kinh H.

Proceedings of the British Machine Vision Conference (BMVC 2011). ed. / Jesse Hoey; Stephen McKenna; Emanuele Trucco. Dundee, United Kingdom : BMVA Press, 2011. p. 1-11.

Research output: A Conference proceeding or a Chapter in BookConference contribution

TY - GEN

T1 - Pose Normalization via Learned 2D Warping for Fully Automatic Face Recognition

AU - Asthana, Arkshay

AU - Jones, Micahel J.

AU - Marks, Tim K.

AU - Goecke, Roland

AU - Tieu, Kinh H.

PY - 2011

Y1 - 2011

N2 - We present a novel approach to pose-invariant face recognition that handles continuous pose variations, is not database-specific, and achieves high accuracy without any manual intervention. Our method uses multidimensional Gaussian process regression to learn a nonlinear mapping function from the 2D shapes of faces at any non-frontal pose to the corresponding 2D frontal face shapes. We use this mapping to take an input image of a new face at an arbitrary pose and pose-normalize it, generating a synthetic frontal image of the face that is then used for recognition. Our fully automatic system for face recognition includes automatic methods for extracting 2D facial feature points and accurately estimating 3D head pose, and this information is used as input to the 2D pose-normalization algorithm. The current system can handle pose variation up to 45 degrees to the left or right (yaw angle) and up to 30 degrees up or down (pitch angle). The system demonstrates high accuracy in recognition experiments on the CMU-PIE, USF 3D, and Multi-PIE databases, showing excellent generalization across databases and convincingly outperforming other automatic methods.

AB - We present a novel approach to pose-invariant face recognition that handles continuous pose variations, is not database-specific, and achieves high accuracy without any manual intervention. Our method uses multidimensional Gaussian process regression to learn a nonlinear mapping function from the 2D shapes of faces at any non-frontal pose to the corresponding 2D frontal face shapes. We use this mapping to take an input image of a new face at an arbitrary pose and pose-normalize it, generating a synthetic frontal image of the face that is then used for recognition. Our fully automatic system for face recognition includes automatic methods for extracting 2D facial feature points and accurately estimating 3D head pose, and this information is used as input to the 2D pose-normalization algorithm. The current system can handle pose variation up to 45 degrees to the left or right (yaw angle) and up to 30 degrees up or down (pitch angle). The system demonstrates high accuracy in recognition experiments on the CMU-PIE, USF 3D, and Multi-PIE databases, showing excellent generalization across databases and convincingly outperforming other automatic methods.

KW - Pose normalization

KW - Face recognition

KW - 2D warping

U2 - 10.5244/C.25.127

DO - 10.5244/C.25.127

M3 - Conference contribution

SN - 978190172543X

SP - 1

EP - 11

BT - Proceedings of the British Machine Vision Conference (BMVC 2011)

A2 - Hoey, Jesse

A2 - McKenna, Stephen

A2 - Trucco, Emanuele

PB - BMVA Press

CY - Dundee, United Kingdom

ER -

Asthana A, Jones MJ, Marks TK, Goecke R, Tieu KH. Pose Normalization via Learned 2D Warping for Fully Automatic Face Recognition. In Hoey J, McKenna S, Trucco E, editors, Proceedings of the British Machine Vision Conference (BMVC 2011). Dundee, United Kingdom: BMVA Press. 2011. p. 1-11 https://doi.org/10.5244/C.25.127