The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
Il existe certaines technologies comme les codes QR pour obtenir des informations numériques à partir d'imprimés. Le filigrane numérique fait partie de ces techniques. Comparé à d’autres techniques, le filigrane numérique permet d’ajouter des informations aux images sans altérer leur conception. À ces fins, des procédés de tatouage numérique pour des imprimés utilisant des marqueurs de détection ou des techniques d'enregistrement d'images pour détecter des zones filigranées sont proposés. Cependant, les marqueurs de détection eux-mêmes peuvent endommager l'apparence, de sorte que les avantages du tatouage numérique, qui ne perd pas en design, ne sont pas pleinement exploités. En revanche, les méthodes utilisant des techniques d'enregistrement d'images ne peuvent pas fonctionner pour des images non enregistrées. Dans cet article, nous proposons une nouvelle méthode de tatouage numérique utilisant l'apprentissage profond pour la détection des zones filigranées au lieu d'utiliser des marqueurs de détection ou l'enregistrement d'images. La méthode proposée introduit une segmentation sémantique basée sur un modèle d'apprentissage profond pour détecter les zones filigranées à partir d'imprimés. Nous préparons deux ensembles de données pour entraîner le modèle d'apprentissage en profondeur. L’une est constituée d’images non filigranées et filigranées transformées géométriquement. Le nombre d'images dans cet ensemble de données est relativement important car les images peuvent être générées sur la base d'un traitement d'image. Cet ensemble de données est utilisé pour la pré-formation. L'autre est obtenu à partir de photographies réellement prises, y compris des imprimés non filigranés ou filigranés. Le nombre de cet ensemble de données est relativement faible car la prise de photographies nécessite beaucoup d'efforts et de temps. Cependant, l'existence de pré-formations permet de réduire le nombre d'images de formation. Cet ensemble de données est utilisé pour un réglage fin afin d'améliorer la robustesse des attaques par caméra d'impression. Lors des expériences, nous avons étudié les performances de notre méthode en la mettant en œuvre sur des smartphones. Les résultats expérimentaux montrent que notre méthode peut transporter 96 bits d’informations avec des imprimés filigranés.
Hiroyuki IMAGAWA
Osaka Prefecture University
Motoi IWATA
Osaka Prefecture University
Koichi KISE
Osaka Prefecture University
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copier
Hiroyuki IMAGAWA, Motoi IWATA, Koichi KISE, "Digital Watermarking Method for Printed Matters Using Deep Learning for Detecting Watermarked Areas" in IEICE TRANSACTIONS on Information,
vol. E104-D, no. 1, pp. 34-42, January 2021, doi: 10.1587/transinf.2020MUP0004.
Abstract: There are some technologies like QR codes to obtain digital information from printed matters. Digital watermarking is one of such techniques. Compared with other techniques, digital watermarking is suitable for adding information to images without spoiling their design. For such purposes, digital watermarking methods for printed matters using detection markers or image registration techniques for detecting watermarked areas are proposed. However, the detection markers themselves can damage the appearance such that the advantages of digital watermarking, which do not lose design, are not fully utilized. On the other hand, methods using image registration techniques are not able to work for non-registered images. In this paper, we propose a novel digital watermarking method using deep learning for the detection of watermarked areas instead of using detection markers or image registration. The proposed method introduces a semantic segmentation based on deep learning model for detecting watermarked areas from printed matters. We prepare two datasets for training the deep learning model. One is constituted of geometrically transformed non-watermarked and watermarked images. The number of images in this dataset is relatively large because the images can be generated based on image processing. This dataset is used for pre-training. The other is obtained from actually taken photographs including non-watermarked or watermarked printed matters. The number of this dataset is relatively small because taking the photographs requires a lot of effort and time. However, the existence of pre-training allows a fewer training images. This dataset is used for fine-tuning to improve robustness for print-cam attacks. In the experiments, we investigated the performance of our method by implementing it on smartphones. The experimental results show that our method can carry 96 bits of information with watermarked printed matters.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2020MUP0004/_p
Copier
@ARTICLE{e104-d_1_34,
author={Hiroyuki IMAGAWA, Motoi IWATA, Koichi KISE, },
journal={IEICE TRANSACTIONS on Information},
title={Digital Watermarking Method for Printed Matters Using Deep Learning for Detecting Watermarked Areas},
year={2021},
volume={E104-D},
number={1},
pages={34-42},
abstract={There are some technologies like QR codes to obtain digital information from printed matters. Digital watermarking is one of such techniques. Compared with other techniques, digital watermarking is suitable for adding information to images without spoiling their design. For such purposes, digital watermarking methods for printed matters using detection markers or image registration techniques for detecting watermarked areas are proposed. However, the detection markers themselves can damage the appearance such that the advantages of digital watermarking, which do not lose design, are not fully utilized. On the other hand, methods using image registration techniques are not able to work for non-registered images. In this paper, we propose a novel digital watermarking method using deep learning for the detection of watermarked areas instead of using detection markers or image registration. The proposed method introduces a semantic segmentation based on deep learning model for detecting watermarked areas from printed matters. We prepare two datasets for training the deep learning model. One is constituted of geometrically transformed non-watermarked and watermarked images. The number of images in this dataset is relatively large because the images can be generated based on image processing. This dataset is used for pre-training. The other is obtained from actually taken photographs including non-watermarked or watermarked printed matters. The number of this dataset is relatively small because taking the photographs requires a lot of effort and time. However, the existence of pre-training allows a fewer training images. This dataset is used for fine-tuning to improve robustness for print-cam attacks. In the experiments, we investigated the performance of our method by implementing it on smartphones. The experimental results show that our method can carry 96 bits of information with watermarked printed matters.},
keywords={},
doi={10.1587/transinf.2020MUP0004},
ISSN={1745-1361},
month={January},}
Copier
TY - JOUR
TI - Digital Watermarking Method for Printed Matters Using Deep Learning for Detecting Watermarked Areas
T2 - IEICE TRANSACTIONS on Information
SP - 34
EP - 42
AU - Hiroyuki IMAGAWA
AU - Motoi IWATA
AU - Koichi KISE
PY - 2021
DO - 10.1587/transinf.2020MUP0004
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E104-D
IS - 1
JA - IEICE TRANSACTIONS on Information
Y1 - January 2021
AB - There are some technologies like QR codes to obtain digital information from printed matters. Digital watermarking is one of such techniques. Compared with other techniques, digital watermarking is suitable for adding information to images without spoiling their design. For such purposes, digital watermarking methods for printed matters using detection markers or image registration techniques for detecting watermarked areas are proposed. However, the detection markers themselves can damage the appearance such that the advantages of digital watermarking, which do not lose design, are not fully utilized. On the other hand, methods using image registration techniques are not able to work for non-registered images. In this paper, we propose a novel digital watermarking method using deep learning for the detection of watermarked areas instead of using detection markers or image registration. The proposed method introduces a semantic segmentation based on deep learning model for detecting watermarked areas from printed matters. We prepare two datasets for training the deep learning model. One is constituted of geometrically transformed non-watermarked and watermarked images. The number of images in this dataset is relatively large because the images can be generated based on image processing. This dataset is used for pre-training. The other is obtained from actually taken photographs including non-watermarked or watermarked printed matters. The number of this dataset is relatively small because taking the photographs requires a lot of effort and time. However, the existence of pre-training allows a fewer training images. This dataset is used for fine-tuning to improve robustness for print-cam attacks. In the experiments, we investigated the performance of our method by implementing it on smartphones. The experimental results show that our method can carry 96 bits of information with watermarked printed matters.
ER -