An Evaluation of Super-Resolution for Face Recognition

Yuxi Peng, Berk Gökberk, Luuk Spreeuwers, Raymond Veldhuis

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

29 Downloads (Pure)

Abstract

We evaluate the performance of face recognition algorithms on images at various resolutions. Then we show to what extent super-resolution (SR) methods can improve the recognition performance when comparing low-resolution (LR) to high-resolution (HR) facial images. Our experiments use both synthetic data (from the FRGC v1.0 database) and surveillance images (from the SCface database). Three face recognition methods are used, namely Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA) and Local Binary Patterns (LBP). Two SR methods are evaluated. The first method learns the mapping between LR images and the corresponding HR images using a regression model. As a result, the reconstructed SR images are close to the HR images that belong to the same subject and far away from others. The second method compares LR and HR facial images without explicitly constructing SR images. It finds a coherent feature space where the correlation of LR and HR is maximum, and then compute the mapping from LR to HR in this feature space. The performance of the two SR methods are compared to that delivered by the standard face recognition without SR. The results show that LDA is mostly robust to resolution changes while LBP is not suitable for the recognition of LR images. SR methods improve the recognition accuracy when downsampled images are used and the first method provides better results than the second one. However, the improvement for realistic LR surveillance images remains limited.
Original languageEnglish
Title of host publication33rd Symposium on Information Theory in the Benelux and the 2nd Joint WIC/IEEE Symposium on Information Theory and Signal Processing in the Benelux
Place of PublicationEnschede
PublisherWerkgemeenschap voor Informatie- en Communicatietheorie (WIC)
Pages36-43
Number of pages8
ISBN (Print)978-90-365-3383-6
Publication statusPublished - 7 Jun 2012
Event33rd WIC Symposium on Information Theory in the Benelux and the 2nd Joint WIC/IEEE Symposium on Information Theory and Signal Processing in the Benelux 2012 - Boekelo, Netherlands
Duration: 24 May 201225 May 2012
Conference number: 33

Conference

Conference33rd WIC Symposium on Information Theory in the Benelux and the 2nd Joint WIC/IEEE Symposium on Information Theory and Signal Processing in the Benelux 2012
CountryNetherlands
CityBoekelo
Period24/05/1225/05/12

Fingerprint

Image resolution
Face recognition
Discriminant analysis
Principal component analysis
Experiments

Keywords

  • METIS-296069
  • IR-80869
  • Face Recognition
  • SCS-Safety
  • Super-resolution
  • Performance Evaluation
  • EWI-22063

Cite this

Peng, Y., Gökberk, B., Spreeuwers, L., & Veldhuis, R. (2012). An Evaluation of Super-Resolution for Face Recognition. In 33rd Symposium on Information Theory in the Benelux and the 2nd Joint WIC/IEEE Symposium on Information Theory and Signal Processing in the Benelux (pp. 36-43). Enschede: Werkgemeenschap voor Informatie- en Communicatietheorie (WIC).
Peng, Yuxi ; Gökberk, Berk ; Spreeuwers, Luuk ; Veldhuis, Raymond. / An Evaluation of Super-Resolution for Face Recognition. 33rd Symposium on Information Theory in the Benelux and the 2nd Joint WIC/IEEE Symposium on Information Theory and Signal Processing in the Benelux. Enschede : Werkgemeenschap voor Informatie- en Communicatietheorie (WIC), 2012. pp. 36-43
@inproceedings{5c3ebaa3a0a44b6b8cba862ebbddf5ac,
title = "An Evaluation of Super-Resolution for Face Recognition",
abstract = "We evaluate the performance of face recognition algorithms on images at various resolutions. Then we show to what extent super-resolution (SR) methods can improve the recognition performance when comparing low-resolution (LR) to high-resolution (HR) facial images. Our experiments use both synthetic data (from the FRGC v1.0 database) and surveillance images (from the SCface database). Three face recognition methods are used, namely Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA) and Local Binary Patterns (LBP). Two SR methods are evaluated. The first method learns the mapping between LR images and the corresponding HR images using a regression model. As a result, the reconstructed SR images are close to the HR images that belong to the same subject and far away from others. The second method compares LR and HR facial images without explicitly constructing SR images. It finds a coherent feature space where the correlation of LR and HR is maximum, and then compute the mapping from LR to HR in this feature space. The performance of the two SR methods are compared to that delivered by the standard face recognition without SR. The results show that LDA is mostly robust to resolution changes while LBP is not suitable for the recognition of LR images. SR methods improve the recognition accuracy when downsampled images are used and the first method provides better results than the second one. However, the improvement for realistic LR surveillance images remains limited.",
keywords = "METIS-296069, IR-80869, Face Recognition, SCS-Safety, Super-resolution, Performance Evaluation, EWI-22063",
author = "Yuxi Peng and Berk G{\"o}kberk and Luuk Spreeuwers and Raymond Veldhuis",
year = "2012",
month = "6",
day = "7",
language = "English",
isbn = "978-90-365-3383-6",
pages = "36--43",
booktitle = "33rd Symposium on Information Theory in the Benelux and the 2nd Joint WIC/IEEE Symposium on Information Theory and Signal Processing in the Benelux",
publisher = "Werkgemeenschap voor Informatie- en Communicatietheorie (WIC)",
address = "Netherlands",

}

Peng, Y, Gökberk, B, Spreeuwers, L & Veldhuis, R 2012, An Evaluation of Super-Resolution for Face Recognition. in 33rd Symposium on Information Theory in the Benelux and the 2nd Joint WIC/IEEE Symposium on Information Theory and Signal Processing in the Benelux. Werkgemeenschap voor Informatie- en Communicatietheorie (WIC), Enschede, pp. 36-43, 33rd WIC Symposium on Information Theory in the Benelux and the 2nd Joint WIC/IEEE Symposium on Information Theory and Signal Processing in the Benelux 2012, Boekelo, Netherlands, 24/05/12.

An Evaluation of Super-Resolution for Face Recognition. / Peng, Yuxi; Gökberk, Berk; Spreeuwers, Luuk; Veldhuis, Raymond.

33rd Symposium on Information Theory in the Benelux and the 2nd Joint WIC/IEEE Symposium on Information Theory and Signal Processing in the Benelux. Enschede : Werkgemeenschap voor Informatie- en Communicatietheorie (WIC), 2012. p. 36-43.

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

TY - GEN

T1 - An Evaluation of Super-Resolution for Face Recognition

AU - Peng, Yuxi

AU - Gökberk, Berk

AU - Spreeuwers, Luuk

AU - Veldhuis, Raymond

PY - 2012/6/7

Y1 - 2012/6/7

N2 - We evaluate the performance of face recognition algorithms on images at various resolutions. Then we show to what extent super-resolution (SR) methods can improve the recognition performance when comparing low-resolution (LR) to high-resolution (HR) facial images. Our experiments use both synthetic data (from the FRGC v1.0 database) and surveillance images (from the SCface database). Three face recognition methods are used, namely Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA) and Local Binary Patterns (LBP). Two SR methods are evaluated. The first method learns the mapping between LR images and the corresponding HR images using a regression model. As a result, the reconstructed SR images are close to the HR images that belong to the same subject and far away from others. The second method compares LR and HR facial images without explicitly constructing SR images. It finds a coherent feature space where the correlation of LR and HR is maximum, and then compute the mapping from LR to HR in this feature space. The performance of the two SR methods are compared to that delivered by the standard face recognition without SR. The results show that LDA is mostly robust to resolution changes while LBP is not suitable for the recognition of LR images. SR methods improve the recognition accuracy when downsampled images are used and the first method provides better results than the second one. However, the improvement for realistic LR surveillance images remains limited.

AB - We evaluate the performance of face recognition algorithms on images at various resolutions. Then we show to what extent super-resolution (SR) methods can improve the recognition performance when comparing low-resolution (LR) to high-resolution (HR) facial images. Our experiments use both synthetic data (from the FRGC v1.0 database) and surveillance images (from the SCface database). Three face recognition methods are used, namely Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA) and Local Binary Patterns (LBP). Two SR methods are evaluated. The first method learns the mapping between LR images and the corresponding HR images using a regression model. As a result, the reconstructed SR images are close to the HR images that belong to the same subject and far away from others. The second method compares LR and HR facial images without explicitly constructing SR images. It finds a coherent feature space where the correlation of LR and HR is maximum, and then compute the mapping from LR to HR in this feature space. The performance of the two SR methods are compared to that delivered by the standard face recognition without SR. The results show that LDA is mostly robust to resolution changes while LBP is not suitable for the recognition of LR images. SR methods improve the recognition accuracy when downsampled images are used and the first method provides better results than the second one. However, the improvement for realistic LR surveillance images remains limited.

KW - METIS-296069

KW - IR-80869

KW - Face Recognition

KW - SCS-Safety

KW - Super-resolution

KW - Performance Evaluation

KW - EWI-22063

M3 - Conference contribution

SN - 978-90-365-3383-6

SP - 36

EP - 43

BT - 33rd Symposium on Information Theory in the Benelux and the 2nd Joint WIC/IEEE Symposium on Information Theory and Signal Processing in the Benelux

PB - Werkgemeenschap voor Informatie- en Communicatietheorie (WIC)

CY - Enschede

ER -

Peng Y, Gökberk B, Spreeuwers L, Veldhuis R. An Evaluation of Super-Resolution for Face Recognition. In 33rd Symposium on Information Theory in the Benelux and the 2nd Joint WIC/IEEE Symposium on Information Theory and Signal Processing in the Benelux. Enschede: Werkgemeenschap voor Informatie- en Communicatietheorie (WIC). 2012. p. 36-43