A Robust Image Watermarking Scheme via Two-Stage Training and Differentiable JPEG Compression
Open accessThe DOI matches its Crossref record; 3 of the 3 references it cites were found in a registry or an index.
Abstract
Digital image watermarking is a vital tool for copyright protection and content authentication. However, most existing methods perform well only under single noise types, while real-world applications often involve composite noises with multiple distortions, leading to poor robustness. To address this issue, we propose a robust image watermarking scheme. To improve performance under combined noise conditions, a two-stage training strategy is introduced: in the first stage, noise intensity increases gradually to stabilize training; in the second stage, mixed strong noises are applied to enhance generalization against complex attacks. Specifically, a strength-balanced watermark optimization algorithm is employed during the testing stage to improve visual quality while maintaining strong robustness. Furthermore, to improve robustness against JPEG compression, we adopt a differentiable fine-grained JPEG module that accurately simulates real compression and enables gradient backpropagation during training. Experimental results demonstrate the superiority of the proposed method under various single and combined distortions. Under noise-free conditions, it achieves 0% bit error rate and 53.55 dB PSNR. Under composite distortions, our scheme maintains a low average BER of 2.40% and a PSNR of 42.70 dB.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source says what the claim attributes to it.
In the citing paper
“Robust watermarking has been considered as a method to embed an authenticity signal directly into media content [6,7].”
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
ImageNet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206848 · Cited by 65,333
ROMark: A Robust Watermarking System Using Adversarial TrainingPreprint
Wen, B., & Aydöre, S. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1910.01221 · Cited by 36
HiDDeN: Hiding Data With Deep NetworksPreprint
Zhu, J., Kaplan, R., Johnson, J. M., et al. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1807.09937 · Cited by 5
Research works printed but not linked: 27
- Hu, 2024, Learning-based image steganography and watermarking: A survey, Expert Syst. Appl., 249, 123715, 10.1016/j.eswa.2024.123715 · DOI
- Hosny, 2024, Digital image watermarking using deep learning: A survey, Comput. Sci. Rev., 53, 100662, 10.1016/j.cosrev.2024.100662 · DOI
- Szepanski, W. (1979, January 16–18). A signal theoretic method for creating forgery-proof documents for automatic verification. Proceedings of the Carnahan Conference on Crime Countermeasures, Lexington, KY, USA.
- Cox, I., Miller, M., Bloom, J., Fridrich, J., and Kalker, T. (2008). Digital Watermarking and Steganography, Morgan Kaufmann Publishers. · DOI · Paper page
- Singh, 2021, Image watermarking using soft computing techniques: A comprehensive survey, Multimed. Tools Appl., 80, 30367, 10.1007/s11042-020-09606-x · doi:10.1007/s11042-020-09606-x
- Mukherjee, 2004, Spatial domain digital watermarking of multimedia objects for buyer authentication, IEEE Trans. Multimed., 6, 1, 10.1109/TMM.2003.819759 · DOI
- Karybali, 2006, Efficient spatial image watermarking via new perceptual masking and blind detection schemes, IEEE Trans. Inf. Forensics Secur., 1, 256, 10.1109/TIFS.2006.873652 · DOI
- Hua, 2020, Informed histogram-based watermarking, IEEE Signal Process. Lett., 27, 236, 10.1109/LSP.2020.2965331 · doi:10.1109/lsp.2020.2965331
- Zhou, 2024, Novel multiple color images encryption and decryption scheme based on a bit-level extension algorithm, Expert Syst. Appl., 238, 122052, 10.1016/j.eswa.2023.122052 · DOI · Paper page
- Hamidi, 2018, Hybrid blind robust image watermarking technique based on DFT-DCT and Arnold transform, Multimed. Tools Appl., 77, 27181, 10.1007/s11042-018-5913-9 · DOI
- Zhou, 2024, A Quantum Image Encryption and Watermarking Algorithm Based on QDCT and Baker map, Int. J. Theor. Phys., 63, 100, 10.1007/s10773-024-05630-x · DOI
- Parah, 2018, Information hiding in edges: A high capacity information hiding technique using hybrid edge detection, Multimed. Tools Appl., 77, 185, 10.1007/s11042-016-4253-x · doi:10.1007/s11042-016-4253-x
- Rabie, 2017, Toward optimal embedding capacity for transform domain steganography: A quad-tree adaptive-region approach, Multimed. Tools Appl., 76, 8627, 10.1007/s11042-016-3501-4 · DOI
- Luo, X., Zhan, R., Chang, H., Yang, F., and Milanfar, P. (2020, January 14–19). Distortion agnostic deep watermarking. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, WA, USA. · DOI · Paper page
- Mun, 2019, Finding robust domain from attacks: A learning framework for blind watermarking, Neurocomputing, 337, 191, 10.1016/j.neucom.2019.01.067 · DOI · Paper page
- Ahmadi, 2020, ReDMark: Framework for residual diffusion watermarking based on deep networks, Expert Syst. Appl., 146, 113157, 10.1016/j.eswa.2019.113157 · DOI · Paper page
- Zhang, 2020, Udh: Universal deep hiding for steganography, watermarking, and light field messaging, Adv. Neural Inf. Process. Syst., 33, 10223
- Liu, Y., Guo, M., Zhang, J., Zhu, Y., and Xie, X. (2019, January 21–25). A novel two-stage separable deep learning framework for practical blind watermarking. Proceedings of the 27th ACM International Conference on Multimedia, Nice, France. · DOI · Paper page
- Jia, Z., Fang, H., and Zhang, W. (2021, January 20–24). Mbrs: Enhancing robustness of dnn-based watermarking by mini-batch of real and simulated jpeg compression. Proceedings of the 29th ACM International Conference on Multimedia, Chengdu, China. · DOI · Paper page
- Wallace, 1991, The JPEG still picture compression standard, Commun. ACM, 34, 30, 10.1145/103085.103089 · DOI · Paper page
- Rumelhart, 1986, Learning representations by back-propagating errors, Nature, 323, 533, 10.1038/323533a0 · DOI · Paper page
- Kandi, 2017, Exploring the learning capabilities of convolutional neural networks for robust image watermarking, Comput. Secur., 65, 247, 10.1016/j.cose.2016.11.016 · DOI · Paper page
- Wang, 2023, Adaptor: Improving the robustness and imperceptibility of watermarking by the adaptive strength factor, IEEE Trans. Circuits Syst. Video Technol., 33, 6260, 10.1109/TCSVT.2023.3265970 · DOI
- Reich, C., Debnath, B., Patel, D., and Chakradhar, S. (March, January 28). Differentiable jpeg: The devil is in the details. Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, Tucson, AZ, USA.
- Shin, R., and Song, D. (2017, January 8). Jpeg-resistant adversarial images. Proceedings of the NIPS 2017 Workshop on Machine Learning and Computer Security, Long Beach, CA, USA.
- Lin, T.Y., Maire, M., Belongie, S., Hays, J., Perona, P., Ramanan, D., Dollár, P., and Zitnick, C.L. (2014, January 6–12). Microsoft coco: Common objects in context. Proceedings of the European Conference on Computer Vision, Zurich, Switzerland. · DOI · Paper page
- Wu, 2025, MCFN: Multi-scale Crossover Feed-forward Network for high performance watermarking, Neurocomputing, 622, 129282, 10.1016/j.neucom.2024.129282 · DOI
Cited by
Yang, C., & Kim, M. (2026) · Applied Sciences · doi:10.3390/app16073391
Metadata
- Authors
- Deng, H., Chen, F., Gan, P., Liao, R., & Yan, X.
- Affiliations
- National University of Defense Technology (CN)
- Corresponding author
- Yan, Xuehu
- Venue
- Electronics
- Publisher
- MDPI AG
- Published
- 2025-11-18
- Volume
- 14
- Issue
- 22
- Pages
- 4510
- ISSN
- 2079-9292 (ISSN-L)
- OpenAlex
- W4416322294
- Cited by
- 1
- References
- 3
- Citations by year
- 2026: 1
- FWCI
- 0.35
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Electronics (publishedVersion, cc-by)
- Funders
- National Natural Science Foundation of China (62271496)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Steganography and Watermarking Techniques
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Data Compression Techniques
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Digital Media Forensic Detection
- Keywords
- image watermarking, robust watermarking, composite noise, two-stage training, copyright protection, content authentication
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.