Super-resolution of facial images in forensics scenarios

Joao Satiro, Kamal Nasrollahi, Paulo Correia, Thomas B. Moeslund

Publikation: Bidrag til bog/antologi/rapport/konference proceedingKonferenceartikel i proceedingForskningpeer review

4 Citationer (Scopus)
402 Downloads (Pure)

Resumé

Forensics facial images are usually provided by
surveillance cameras and are therefore of poor quality and resolution.
Simple upsampling algorithms can not produce artifact-free
higher resolution images from such low-resolution (LR) images.
To deal with that, reconstruction-based super-resolution (SR)
algorithms might be used. But, the problem with these algorithms
is that they mostly require motion estimation between LR and
low-quality images which is not always practical. To deal with
this, we first simply interpolate the LR input images and then
perform motion estimation. The estimated motion parameters are
then used in a non-local mean-based SR algorithm to produce
a higher quality image. This image is further fused with the
interpolated version of the reference image via an alpha-blending
approach. The experimental results on benchmark datasets and
locally collected videos from surveillance cameras, show the outperformance of the proposed system over similar ones.
OriginalsprogEngelsk
TitelIEEE 5th International Conference on Image Processing Theory, Tools and Applications
ForlagIEEE Signal Processing Society
Publikationsdato2015
Sider55-60
ISBN (Trykt)978-1-4799-8636-1, 978-1-4799-8637-8
DOI
StatusUdgivet - 2015
BegivenhedIEEE International Conference on Image Processing Theory, Tools and Applications - Orleans, Frankrig
Varighed: 10 nov. 201513 nov. 2015
Konferencens nummer: 5th

Konference

KonferenceIEEE International Conference on Image Processing Theory, Tools and Applications
Nummer5th
LandFrankrig
ByOrleans
Periode10/11/201513/11/2015

Fingerprint

Motion estimation
Image resolution
Image quality
Cameras

Citer dette

Satiro, J., Nasrollahi, K., Correia, P., & Moeslund, T. B. (2015). Super-resolution of facial images in forensics scenarios. I IEEE 5th International Conference on Image Processing Theory, Tools and Applications (s. 55-60). IEEE Signal Processing Society. https://doi.org/10.1109/IPTA.2015.7367096
Satiro, Joao ; Nasrollahi, Kamal ; Correia, Paulo ; Moeslund, Thomas B. / Super-resolution of facial images in forensics scenarios. IEEE 5th International Conference on Image Processing Theory, Tools and Applications. IEEE Signal Processing Society, 2015. s. 55-60
@inproceedings{c5649c3cca184a32a4e7a26667c60c74,
title = "Super-resolution of facial images in forensics scenarios",
abstract = "Forensics facial images are usually provided bysurveillance cameras and are therefore of poor quality and resolution.Simple upsampling algorithms can not produce artifact-freehigher resolution images from such low-resolution (LR) images.To deal with that, reconstruction-based super-resolution (SR)algorithms might be used. But, the problem with these algorithmsis that they mostly require motion estimation between LR andlow-quality images which is not always practical. To deal withthis, we first simply interpolate the LR input images and thenperform motion estimation. The estimated motion parameters arethen used in a non-local mean-based SR algorithm to producea higher quality image. This image is further fused with theinterpolated version of the reference image via an alpha-blendingapproach. The experimental results on benchmark datasets andlocally collected videos from surveillance cameras, show the outperformance of the proposed system over similar ones.",
keywords = "super-resolution, reconstruction, facial images, Forensics",
author = "Joao Satiro and Kamal Nasrollahi and Paulo Correia and Moeslund, {Thomas B.}",
year = "2015",
doi = "10.1109/IPTA.2015.7367096",
language = "English",
isbn = "978-1-4799-8636-1",
pages = "55--60",
booktitle = "IEEE 5th International Conference on Image Processing Theory, Tools and Applications",
publisher = "IEEE Signal Processing Society",
address = "United States",

}

Satiro, J, Nasrollahi, K, Correia, P & Moeslund, TB 2015, Super-resolution of facial images in forensics scenarios. i IEEE 5th International Conference on Image Processing Theory, Tools and Applications. IEEE Signal Processing Society, s. 55-60, IEEE International Conference on Image Processing Theory, Tools and Applications , Orleans, Frankrig, 10/11/2015. https://doi.org/10.1109/IPTA.2015.7367096

Super-resolution of facial images in forensics scenarios. / Satiro, Joao; Nasrollahi, Kamal; Correia, Paulo; Moeslund, Thomas B.

IEEE 5th International Conference on Image Processing Theory, Tools and Applications. IEEE Signal Processing Society, 2015. s. 55-60.

Publikation: Bidrag til bog/antologi/rapport/konference proceedingKonferenceartikel i proceedingForskningpeer review

TY - GEN

T1 - Super-resolution of facial images in forensics scenarios

AU - Satiro, Joao

AU - Nasrollahi, Kamal

AU - Correia, Paulo

AU - Moeslund, Thomas B.

PY - 2015

Y1 - 2015

N2 - Forensics facial images are usually provided bysurveillance cameras and are therefore of poor quality and resolution.Simple upsampling algorithms can not produce artifact-freehigher resolution images from such low-resolution (LR) images.To deal with that, reconstruction-based super-resolution (SR)algorithms might be used. But, the problem with these algorithmsis that they mostly require motion estimation between LR andlow-quality images which is not always practical. To deal withthis, we first simply interpolate the LR input images and thenperform motion estimation. The estimated motion parameters arethen used in a non-local mean-based SR algorithm to producea higher quality image. This image is further fused with theinterpolated version of the reference image via an alpha-blendingapproach. The experimental results on benchmark datasets andlocally collected videos from surveillance cameras, show the outperformance of the proposed system over similar ones.

AB - Forensics facial images are usually provided bysurveillance cameras and are therefore of poor quality and resolution.Simple upsampling algorithms can not produce artifact-freehigher resolution images from such low-resolution (LR) images.To deal with that, reconstruction-based super-resolution (SR)algorithms might be used. But, the problem with these algorithmsis that they mostly require motion estimation between LR andlow-quality images which is not always practical. To deal withthis, we first simply interpolate the LR input images and thenperform motion estimation. The estimated motion parameters arethen used in a non-local mean-based SR algorithm to producea higher quality image. This image is further fused with theinterpolated version of the reference image via an alpha-blendingapproach. The experimental results on benchmark datasets andlocally collected videos from surveillance cameras, show the outperformance of the proposed system over similar ones.

KW - super-resolution

KW - reconstruction

KW - facial images

KW - Forensics

UR - http://ipta.fr/ipta15/index.php/program

U2 - 10.1109/IPTA.2015.7367096

DO - 10.1109/IPTA.2015.7367096

M3 - Article in proceeding

SN - 978-1-4799-8636-1

SN - 978-1-4799-8637-8

SP - 55

EP - 60

BT - IEEE 5th International Conference on Image Processing Theory, Tools and Applications

PB - IEEE Signal Processing Society

ER -

Satiro J, Nasrollahi K, Correia P, Moeslund TB. Super-resolution of facial images in forensics scenarios. I IEEE 5th International Conference on Image Processing Theory, Tools and Applications. IEEE Signal Processing Society. 2015. s. 55-60 https://doi.org/10.1109/IPTA.2015.7367096