Face recognition at a distance: low-resolution and alignment problems

Yuxi Peng

Research output: ThesisPhD Thesis - Research UT, graduation UTAcademic

51 Downloads (Pure)

Abstract

Existing face recognition techniques are very successful in recognizing high-resolution facial images. However, their performance is not sufficient on low-resolution facial images. In this thesis we focus on dealing with a typical forensic scenario, where the gallery images are the high-resolution mug-shots in the police database and the probe images are the low-resolution surveillance quality facial images.

We proposed a novel method, mixed-resolution biometric comparison, which allows direct low-resolution to high-resolution comparison. The method is based on the likelihood ratio framework where in the derivation of the expression for the likelihood ratio, the combined statistics of the low- and high-resolution images is taken into account. Our experiments on surveillance quality images demonstrate that this method significantly outperforms the state-of-the-art.

In literature on low-resolution face recognition, what in some papers is considered as low-resolution, is still considered as high-resolution in other papers. To harmonize the terminology in low-resolution face recognition, we propose a resolution scale. We define the range of low-resolution and further divide it into Upper Low Resolution, Moderately Low Resolution and Very Low Resolution.

Because the lack of low-resolution images, most of the existing low-resolution face recognition methods are trained and tested using down-sampled images. In this thesis, we test various face recognition methods and demonstrate that down-sampled images are not fully representative of realistic low-resolution images. We further demonstrate that, inaccurate alignment is the major problem that causes the poor recognition performance on real low-resolution images. In addition, we propose to use matching-score based registration to achieve better alignment and hence better face recognition performance.

In conclusion, we propose solutions to compare low-resolution probes with high-resolution galleries which significantly outperform the state-of-the-art on surveillance quality facial images. We emphasise that realistic low-resolution material should be used for training and testing. We focus attention on developing face recognition methods that can actually be useful for real-life applications. We bring an important step forward of low-resolution face recognition for forensic search.
Original languageEnglish
QualificationDoctor of Philosophy
Awarding Institution
  • University of Twente
Supervisors/Advisors
  • Veldhuis, Raymond N.J., Supervisor
  • Spreeuwers, Lieuwe Jan, Supervisor
Award date8 Feb 2019
Place of PublicationEnschede
Publisher
Print ISBNs978-90-365-4711-6
DOIs
Publication statusPublished - 8 Feb 2019

Fingerprint

Face recognition
Image resolution
Optical resolving power
Law enforcement
Biometrics
Terminology
Image quality
Statistics
Testing

Cite this

Peng, Yuxi. / Face recognition at a distance : low-resolution and alignment problems. Enschede : University of Twente, 2019. 131 p.
@phdthesis{1d6a3d288b444063bb61b4ad74977365,
title = "Face recognition at a distance: low-resolution and alignment problems",
abstract = "Existing face recognition techniques are very successful in recognizing high-resolution facial images. However, their performance is not sufficient on low-resolution facial images. In this thesis we focus on dealing with a typical forensic scenario, where the gallery images are the high-resolution mug-shots in the police database and the probe images are the low-resolution surveillance quality facial images. We proposed a novel method, mixed-resolution biometric comparison, which allows direct low-resolution to high-resolution comparison. The method is based on the likelihood ratio framework where in the derivation of the expression for the likelihood ratio, the combined statistics of the low- and high-resolution images is taken into account. Our experiments on surveillance quality images demonstrate that this method significantly outperforms the state-of-the-art. In literature on low-resolution face recognition, what in some papers is considered as low-resolution, is still considered as high-resolution in other papers. To harmonize the terminology in low-resolution face recognition, we propose a resolution scale. We define the range of low-resolution and further divide it into Upper Low Resolution, Moderately Low Resolution and Very Low Resolution. Because the lack of low-resolution images, most of the existing low-resolution face recognition methods are trained and tested using down-sampled images. In this thesis, we test various face recognition methods and demonstrate that down-sampled images are not fully representative of realistic low-resolution images. We further demonstrate that, inaccurate alignment is the major problem that causes the poor recognition performance on real low-resolution images. In addition, we propose to use matching-score based registration to achieve better alignment and hence better face recognition performance.In conclusion, we propose solutions to compare low-resolution probes with high-resolution galleries which significantly outperform the state-of-the-art on surveillance quality facial images. We emphasise that realistic low-resolution material should be used for training and testing. We focus attention on developing face recognition methods that can actually be useful for real-life applications. We bring an important step forward of low-resolution face recognition for forensic search.",
author = "Yuxi Peng",
year = "2019",
month = "2",
day = "8",
doi = "10.3990/1.9789036547116",
language = "English",
isbn = "978-90-365-4711-6",
publisher = "University of Twente",
address = "Netherlands",
school = "University of Twente",

}

Peng, Y 2019, 'Face recognition at a distance: low-resolution and alignment problems', Doctor of Philosophy, University of Twente, Enschede. https://doi.org/10.3990/1.9789036547116

Face recognition at a distance : low-resolution and alignment problems. / Peng, Yuxi.

Enschede : University of Twente, 2019. 131 p.

Research output: ThesisPhD Thesis - Research UT, graduation UTAcademic

TY - THES

T1 - Face recognition at a distance

T2 - low-resolution and alignment problems

AU - Peng, Yuxi

PY - 2019/2/8

Y1 - 2019/2/8

N2 - Existing face recognition techniques are very successful in recognizing high-resolution facial images. However, their performance is not sufficient on low-resolution facial images. In this thesis we focus on dealing with a typical forensic scenario, where the gallery images are the high-resolution mug-shots in the police database and the probe images are the low-resolution surveillance quality facial images. We proposed a novel method, mixed-resolution biometric comparison, which allows direct low-resolution to high-resolution comparison. The method is based on the likelihood ratio framework where in the derivation of the expression for the likelihood ratio, the combined statistics of the low- and high-resolution images is taken into account. Our experiments on surveillance quality images demonstrate that this method significantly outperforms the state-of-the-art. In literature on low-resolution face recognition, what in some papers is considered as low-resolution, is still considered as high-resolution in other papers. To harmonize the terminology in low-resolution face recognition, we propose a resolution scale. We define the range of low-resolution and further divide it into Upper Low Resolution, Moderately Low Resolution and Very Low Resolution. Because the lack of low-resolution images, most of the existing low-resolution face recognition methods are trained and tested using down-sampled images. In this thesis, we test various face recognition methods and demonstrate that down-sampled images are not fully representative of realistic low-resolution images. We further demonstrate that, inaccurate alignment is the major problem that causes the poor recognition performance on real low-resolution images. In addition, we propose to use matching-score based registration to achieve better alignment and hence better face recognition performance.In conclusion, we propose solutions to compare low-resolution probes with high-resolution galleries which significantly outperform the state-of-the-art on surveillance quality facial images. We emphasise that realistic low-resolution material should be used for training and testing. We focus attention on developing face recognition methods that can actually be useful for real-life applications. We bring an important step forward of low-resolution face recognition for forensic search.

AB - Existing face recognition techniques are very successful in recognizing high-resolution facial images. However, their performance is not sufficient on low-resolution facial images. In this thesis we focus on dealing with a typical forensic scenario, where the gallery images are the high-resolution mug-shots in the police database and the probe images are the low-resolution surveillance quality facial images. We proposed a novel method, mixed-resolution biometric comparison, which allows direct low-resolution to high-resolution comparison. The method is based on the likelihood ratio framework where in the derivation of the expression for the likelihood ratio, the combined statistics of the low- and high-resolution images is taken into account. Our experiments on surveillance quality images demonstrate that this method significantly outperforms the state-of-the-art. In literature on low-resolution face recognition, what in some papers is considered as low-resolution, is still considered as high-resolution in other papers. To harmonize the terminology in low-resolution face recognition, we propose a resolution scale. We define the range of low-resolution and further divide it into Upper Low Resolution, Moderately Low Resolution and Very Low Resolution. Because the lack of low-resolution images, most of the existing low-resolution face recognition methods are trained and tested using down-sampled images. In this thesis, we test various face recognition methods and demonstrate that down-sampled images are not fully representative of realistic low-resolution images. We further demonstrate that, inaccurate alignment is the major problem that causes the poor recognition performance on real low-resolution images. In addition, we propose to use matching-score based registration to achieve better alignment and hence better face recognition performance.In conclusion, we propose solutions to compare low-resolution probes with high-resolution galleries which significantly outperform the state-of-the-art on surveillance quality facial images. We emphasise that realistic low-resolution material should be used for training and testing. We focus attention on developing face recognition methods that can actually be useful for real-life applications. We bring an important step forward of low-resolution face recognition for forensic search.

U2 - 10.3990/1.9789036547116

DO - 10.3990/1.9789036547116

M3 - PhD Thesis - Research UT, graduation UT

SN - 978-90-365-4711-6

PB - University of Twente

CY - Enschede

ER -