Initiating GrabCut by Color Difference for Automatic Foreground Extraction of Passport Imagery

Adriá Arbués Sangüesa, Nicolai Krogh Jørgensen, Christian Aagaard Larsen, Kamal Nasrollahi, Thomas B. Moeslund

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

1 Citation (Scopus)
566 Downloads (Pure)

Resumé

Grabcut, an iterative algorithm based on Graph Cut, is a popular foreground segmentation method. However, it suffers from a main drawback: a manual interaction is required in order to start segmenting the image. In this paper, four different methods based on image pairs are used to obtain an initial extraction of the foreground. Then, the obtained initial estimation of the foreground is used as input to the GrabCut algorithm, thus avoiding the need of interaction. Moreover, this paper is focused on passport images, which require an almost pixel-perfect segmentation in order to be a valid photo. Having gathered our own dataset and generated ground truth images, promising results are obtained in terms of F1-scores, with a maximum mean of 0.975 among all the images, improving the performance of GrabCut in all cases. Some future work directions are given for those unsolved issues that were faced, such as the segmentation in hair regions or tests in a non-uniform background scenario.
OriginalsprogEngelsk
TitelIEEE International Conference on Image Processing Theory, Tools and Applications
ForlagIEEE
Publikationsdatodec. 2016
ISBN (Trykt)978-1-4673-8911-2
ISBN (Elektronisk)978-1-4673-8910-5
DOI
StatusUdgivet - dec. 2016
BegivenhedIEEE International Conference on Image Processing Theory, Tools and Applications - Oulu, Finland
Varighed: 12 dec. 201615 dec. 2016
Konferencens nummer: 6
http://www.ipta-conference.com/ipta16/

Konference

KonferenceIEEE International Conference on Image Processing Theory, Tools and Applications
Nummer6
LandFinland
ByOulu
Periode12/12/201615/12/2016
Internetadresse

Fingerprint

Color
Pixels

Citer dette

Sangüesa, A. A., Jørgensen, N. K., Larsen, C. A., Nasrollahi, K., & Moeslund, T. B. (2016). Initiating GrabCut by Color Difference for Automatic Foreground Extraction of Passport Imagery. I IEEE International Conference on Image Processing Theory, Tools and Applications IEEE. https://doi.org/10.1109/IPTA.2016.7820964
Sangüesa, Adriá Arbués ; Jørgensen, Nicolai Krogh ; Larsen, Christian Aagaard ; Nasrollahi, Kamal ; Moeslund, Thomas B. / Initiating GrabCut by Color Difference for Automatic Foreground Extraction of Passport Imagery. IEEE International Conference on Image Processing Theory, Tools and Applications. IEEE, 2016.
@inproceedings{67e3f1d0530a4a56a41400ffd2edd8e7,
title = "Initiating GrabCut by Color Difference for Automatic Foreground Extraction of Passport Imagery",
abstract = "Grabcut, an iterative algorithm based on Graph Cut, is a popular foreground segmentation method. However, it suffers from a main drawback: a manual interaction is required in order to start segmenting the image. In this paper, four different methods based on image pairs are used to obtain an initial extraction of the foreground. Then, the obtained initial estimation of the foreground is used as input to the GrabCut algorithm, thus avoiding the need of interaction. Moreover, this paper is focused on passport images, which require an almost pixel-perfect segmentation in order to be a valid photo. Having gathered our own dataset and generated ground truth images, promising results are obtained in terms of F1-scores, with a maximum mean of 0.975 among all the images, improving the performance of GrabCut in all cases. Some future work directions are given for those unsolved issues that were faced, such as the segmentation in hair regions or tests in a non-uniform background scenario.",
keywords = "Foreground Extraction, Image Pairs, GrabCut, Passport Imagery, Color Difference",
author = "Sang{\"u}esa, {Adri{\'a} Arbu{\'e}s} and J{\o}rgensen, {Nicolai Krogh} and Larsen, {Christian Aagaard} and Kamal Nasrollahi and Moeslund, {Thomas B.}",
year = "2016",
month = "12",
doi = "10.1109/IPTA.2016.7820964",
language = "English",
isbn = "978-1-4673-8911-2",
booktitle = "IEEE International Conference on Image Processing Theory, Tools and Applications",
publisher = "IEEE",
address = "United States",

}

Sangüesa, AA, Jørgensen, NK, Larsen, CA, Nasrollahi, K & Moeslund, TB 2016, Initiating GrabCut by Color Difference for Automatic Foreground Extraction of Passport Imagery. i IEEE International Conference on Image Processing Theory, Tools and Applications. IEEE, IEEE International Conference on Image Processing Theory, Tools and Applications, Oulu, Finland, 12/12/2016. https://doi.org/10.1109/IPTA.2016.7820964

Initiating GrabCut by Color Difference for Automatic Foreground Extraction of Passport Imagery. / Sangüesa, Adriá Arbués; Jørgensen, Nicolai Krogh; Larsen, Christian Aagaard; Nasrollahi, Kamal; Moeslund, Thomas B.

IEEE International Conference on Image Processing Theory, Tools and Applications. IEEE, 2016.

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

TY - GEN

T1 - Initiating GrabCut by Color Difference for Automatic Foreground Extraction of Passport Imagery

AU - Sangüesa, Adriá Arbués

AU - Jørgensen, Nicolai Krogh

AU - Larsen, Christian Aagaard

AU - Nasrollahi, Kamal

AU - Moeslund, Thomas B.

PY - 2016/12

Y1 - 2016/12

N2 - Grabcut, an iterative algorithm based on Graph Cut, is a popular foreground segmentation method. However, it suffers from a main drawback: a manual interaction is required in order to start segmenting the image. In this paper, four different methods based on image pairs are used to obtain an initial extraction of the foreground. Then, the obtained initial estimation of the foreground is used as input to the GrabCut algorithm, thus avoiding the need of interaction. Moreover, this paper is focused on passport images, which require an almost pixel-perfect segmentation in order to be a valid photo. Having gathered our own dataset and generated ground truth images, promising results are obtained in terms of F1-scores, with a maximum mean of 0.975 among all the images, improving the performance of GrabCut in all cases. Some future work directions are given for those unsolved issues that were faced, such as the segmentation in hair regions or tests in a non-uniform background scenario.

AB - Grabcut, an iterative algorithm based on Graph Cut, is a popular foreground segmentation method. However, it suffers from a main drawback: a manual interaction is required in order to start segmenting the image. In this paper, four different methods based on image pairs are used to obtain an initial extraction of the foreground. Then, the obtained initial estimation of the foreground is used as input to the GrabCut algorithm, thus avoiding the need of interaction. Moreover, this paper is focused on passport images, which require an almost pixel-perfect segmentation in order to be a valid photo. Having gathered our own dataset and generated ground truth images, promising results are obtained in terms of F1-scores, with a maximum mean of 0.975 among all the images, improving the performance of GrabCut in all cases. Some future work directions are given for those unsolved issues that were faced, such as the segmentation in hair regions or tests in a non-uniform background scenario.

KW - Foreground Extraction

KW - Image Pairs

KW - GrabCut

KW - Passport Imagery

KW - Color Difference

U2 - 10.1109/IPTA.2016.7820964

DO - 10.1109/IPTA.2016.7820964

M3 - Article in proceeding

SN - 978-1-4673-8911-2

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

PB - IEEE

ER -

Sangüesa AA, Jørgensen NK, Larsen CA, Nasrollahi K, Moeslund TB. Initiating GrabCut by Color Difference for Automatic Foreground Extraction of Passport Imagery. I IEEE International Conference on Image Processing Theory, Tools and Applications. IEEE. 2016 https://doi.org/10.1109/IPTA.2016.7820964