Comparison of super-resolution benefits for downsampled iages and real low-resolution data

Yuxi Peng, Luuk Spreeuwers, Bert Gökberk, Raymond Veldhuis

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademic

7 Downloads (Pure)

Abstract

Recently, more and more researchers are exploring the benefits of super-resolution methods on low-resolution face recognition. However, often results presented are obtained on downsampled high-resolution face images. Because downsampled images are different from real images taken at low resolution, it is important to include real surveillance data. In this paper, we investigate the difference between downsampled images and real surveillance data in two aspects: (1) the influence of resolution on face recognition accuracy, and (2) the improvement of accuracy that can be achieved by super-resolution on these images. Specifically, we will test the following hypotheses: (1) face recognition performance on real images is much worse than on downsampled images, and (2) super-resolution improves the performance of downsampled images more than real images. Our experiments are conducted using videos from the HumanID database. In each video, the target person’s face is captured while he is walking towards the surveillance camera. We detect the faces in the video frames using a Viola-Jones face detector. Then we select face images of four different resolutions: two low-resolution and two high-resolution. The high-resolution images are used for gallery and generating downsampled images. We perform two types of face recognition experiments. In the first type of experiments, three face recognition methods are evaluated for images with different resolution. The three methods are (1) Principal Component Analysis, (2) Linear Discriminant Analysis, and (3) Local Binary Patterns. In the second type, we apply two super-resolution methods: (1) a model based method and (2) a feature based method on the low-resolution (both real and downsampled) images and then compute the face recognition accuracy.
Original languageEnglish
Title of host publication34th WIC Symposium on Information Theory in the Benelux and the 3rd Joint WIC/IEEE Symposium on Information Theory and Signal Processing in the Benelux 2013
Subtitle of host publicationLeuven, Belgium, 30-31 May 2013
PublisherWerkgemeenschap voor Informatie- en Communicatietheorie (WIC)
Pages244-251
Number of pages8
ISBN (Print)9781627487375, 978-90-365-0000-5
Publication statusPublished - 30 May 2013
Event34th WIC Symposium on Information Theory in the Benelux 2013 - Leuven, Belgium
Duration: 30 May 201331 May 2013
Conference number: 34

Conference

Conference34th WIC Symposium on Information Theory in the Benelux 2013
CountryBelgium
CityLeuven
Period30/05/1331/05/13

Fingerprint

Face recognition
Optical resolving power
Experiments
Discriminant analysis
Image resolution
Principal component analysis
Cameras
Detectors

Keywords

  • EWI-23950
  • SCS-Safety
  • Super-resolution
  • IR-87825
  • Face Recognition
  • METIS-300149
  • Performance Evaluation

Cite this

Peng, Y., Spreeuwers, L., Gökberk, B., & Veldhuis, R. (2013). Comparison of super-resolution benefits for downsampled iages and real low-resolution data. In 34th WIC Symposium on Information Theory in the Benelux and the 3rd Joint WIC/IEEE Symposium on Information Theory and Signal Processing in the Benelux 2013: Leuven, Belgium, 30-31 May 2013 (pp. 244-251). Werkgemeenschap voor Informatie- en Communicatietheorie (WIC).
Peng, Yuxi ; Spreeuwers, Luuk ; Gökberk, Bert ; Veldhuis, Raymond. / Comparison of super-resolution benefits for downsampled iages and real low-resolution data. 34th WIC Symposium on Information Theory in the Benelux and the 3rd Joint WIC/IEEE Symposium on Information Theory and Signal Processing in the Benelux 2013: Leuven, Belgium, 30-31 May 2013. Werkgemeenschap voor Informatie- en Communicatietheorie (WIC), 2013. pp. 244-251
@inproceedings{fe2cca4253ec47a18d7c8de17bbe2547,
title = "Comparison of super-resolution benefits for downsampled iages and real low-resolution data",
abstract = "Recently, more and more researchers are exploring the benefits of super-resolution methods on low-resolution face recognition. However, often results presented are obtained on downsampled high-resolution face images. Because downsampled images are different from real images taken at low resolution, it is important to include real surveillance data. In this paper, we investigate the difference between downsampled images and real surveillance data in two aspects: (1) the influence of resolution on face recognition accuracy, and (2) the improvement of accuracy that can be achieved by super-resolution on these images. Specifically, we will test the following hypotheses: (1) face recognition performance on real images is much worse than on downsampled images, and (2) super-resolution improves the performance of downsampled images more than real images. Our experiments are conducted using videos from the HumanID database. In each video, the target person’s face is captured while he is walking towards the surveillance camera. We detect the faces in the video frames using a Viola-Jones face detector. Then we select face images of four different resolutions: two low-resolution and two high-resolution. The high-resolution images are used for gallery and generating downsampled images. We perform two types of face recognition experiments. In the first type of experiments, three face recognition methods are evaluated for images with different resolution. The three methods are (1) Principal Component Analysis, (2) Linear Discriminant Analysis, and (3) Local Binary Patterns. In the second type, we apply two super-resolution methods: (1) a model based method and (2) a feature based method on the low-resolution (both real and downsampled) images and then compute the face recognition accuracy.",
keywords = "EWI-23950, SCS-Safety, Super-resolution, IR-87825, Face Recognition, METIS-300149, Performance Evaluation",
author = "Yuxi Peng and Luuk Spreeuwers and Bert G{\"o}kberk and Raymond Veldhuis",
year = "2013",
month = "5",
day = "30",
language = "English",
isbn = "9781627487375",
pages = "244--251",
booktitle = "34th WIC Symposium on Information Theory in the Benelux and the 3rd Joint WIC/IEEE Symposium on Information Theory and Signal Processing in the Benelux 2013",
publisher = "Werkgemeenschap voor Informatie- en Communicatietheorie (WIC)",
address = "Netherlands",

}

Peng, Y, Spreeuwers, L, Gökberk, B & Veldhuis, R 2013, Comparison of super-resolution benefits for downsampled iages and real low-resolution data. in 34th WIC Symposium on Information Theory in the Benelux and the 3rd Joint WIC/IEEE Symposium on Information Theory and Signal Processing in the Benelux 2013: Leuven, Belgium, 30-31 May 2013. Werkgemeenschap voor Informatie- en Communicatietheorie (WIC), pp. 244-251, 34th WIC Symposium on Information Theory in the Benelux 2013, Leuven, Belgium, 30/05/13.

Comparison of super-resolution benefits for downsampled iages and real low-resolution data. / Peng, Yuxi; Spreeuwers, Luuk; Gökberk, Bert; Veldhuis, Raymond.

34th WIC Symposium on Information Theory in the Benelux and the 3rd Joint WIC/IEEE Symposium on Information Theory and Signal Processing in the Benelux 2013: Leuven, Belgium, 30-31 May 2013. Werkgemeenschap voor Informatie- en Communicatietheorie (WIC), 2013. p. 244-251.

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademic

TY - GEN

T1 - Comparison of super-resolution benefits for downsampled iages and real low-resolution data

AU - Peng, Yuxi

AU - Spreeuwers, Luuk

AU - Gökberk, Bert

AU - Veldhuis, Raymond

PY - 2013/5/30

Y1 - 2013/5/30

N2 - Recently, more and more researchers are exploring the benefits of super-resolution methods on low-resolution face recognition. However, often results presented are obtained on downsampled high-resolution face images. Because downsampled images are different from real images taken at low resolution, it is important to include real surveillance data. In this paper, we investigate the difference between downsampled images and real surveillance data in two aspects: (1) the influence of resolution on face recognition accuracy, and (2) the improvement of accuracy that can be achieved by super-resolution on these images. Specifically, we will test the following hypotheses: (1) face recognition performance on real images is much worse than on downsampled images, and (2) super-resolution improves the performance of downsampled images more than real images. Our experiments are conducted using videos from the HumanID database. In each video, the target person’s face is captured while he is walking towards the surveillance camera. We detect the faces in the video frames using a Viola-Jones face detector. Then we select face images of four different resolutions: two low-resolution and two high-resolution. The high-resolution images are used for gallery and generating downsampled images. We perform two types of face recognition experiments. In the first type of experiments, three face recognition methods are evaluated for images with different resolution. The three methods are (1) Principal Component Analysis, (2) Linear Discriminant Analysis, and (3) Local Binary Patterns. In the second type, we apply two super-resolution methods: (1) a model based method and (2) a feature based method on the low-resolution (both real and downsampled) images and then compute the face recognition accuracy.

AB - Recently, more and more researchers are exploring the benefits of super-resolution methods on low-resolution face recognition. However, often results presented are obtained on downsampled high-resolution face images. Because downsampled images are different from real images taken at low resolution, it is important to include real surveillance data. In this paper, we investigate the difference between downsampled images and real surveillance data in two aspects: (1) the influence of resolution on face recognition accuracy, and (2) the improvement of accuracy that can be achieved by super-resolution on these images. Specifically, we will test the following hypotheses: (1) face recognition performance on real images is much worse than on downsampled images, and (2) super-resolution improves the performance of downsampled images more than real images. Our experiments are conducted using videos from the HumanID database. In each video, the target person’s face is captured while he is walking towards the surveillance camera. We detect the faces in the video frames using a Viola-Jones face detector. Then we select face images of four different resolutions: two low-resolution and two high-resolution. The high-resolution images are used for gallery and generating downsampled images. We perform two types of face recognition experiments. In the first type of experiments, three face recognition methods are evaluated for images with different resolution. The three methods are (1) Principal Component Analysis, (2) Linear Discriminant Analysis, and (3) Local Binary Patterns. In the second type, we apply two super-resolution methods: (1) a model based method and (2) a feature based method on the low-resolution (both real and downsampled) images and then compute the face recognition accuracy.

KW - EWI-23950

KW - SCS-Safety

KW - Super-resolution

KW - IR-87825

KW - Face Recognition

KW - METIS-300149

KW - Performance Evaluation

M3 - Conference contribution

SN - 9781627487375

SN - 978-90-365-0000-5

SP - 244

EP - 251

BT - 34th WIC Symposium on Information Theory in the Benelux and the 3rd Joint WIC/IEEE Symposium on Information Theory and Signal Processing in the Benelux 2013

PB - Werkgemeenschap voor Informatie- en Communicatietheorie (WIC)

ER -

Peng Y, Spreeuwers L, Gökberk B, Veldhuis R. Comparison of super-resolution benefits for downsampled iages and real low-resolution data. In 34th WIC Symposium on Information Theory in the Benelux and the 3rd Joint WIC/IEEE Symposium on Information Theory and Signal Processing in the Benelux 2013: Leuven, Belgium, 30-31 May 2013. Werkgemeenschap voor Informatie- en Communicatietheorie (WIC). 2013. p. 244-251