SSH-Net: A Self-Supervised and Hybrid Network for Noisy Image Watermark Removal

Jianjun Gao; Kim-Hui Yap; Wenyang Liu

REVIEW 3 major objections 7 minor 42 references

SSH-Net: A Self-Supervised and Hybrid Network for Noisy Image Watermark Removal

T0 review · 3 major / 7 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Watermark removal can be learned from the watermarked image alone, without paired clean examples, by treating randomly re-watermarked copies as the training target.

desk verdict Plausible architecture, extensive experiments, but the self-supervised target construction contradicts the paper's own blend equation; the PSNR gains are unexplained. read the letter →

arxiv 2505.05088 v1 pith:GEJJN7D6 submitted 2025-05-08 cs.MM cs.CVeess.IV

Wenyang Liu , Jianjun Gao , Kim-Hui Yap This is my paper · ORCID

classification cs.MMcs.CVeess.IV

keywords self-supervisedlearningvisiblewatermarkremovalimagerestorationdenoisingvisiontransformersparseself-attentionnoisy

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

SSH-Net claims that visible watermark removal from noisy images can be trained without paired clean images. The paper synthesizes a reference image by randomly adding more watermarks to an already-watermarked input, asserts that this random target is an unbiased estimate of the clean image in expectation, and trains the network against that target. If correct, the method removes the main practical obstacle to watermark removal: real-world images seldom come with clean counterparts. The paper reports that the resulting hybrid CNN-Transformer network exceeds previous methods on standard noisy-watermark benchmarks while using fewer FLOPs.

What carries the argument

The load-bearing object is the synthesized target $Y_w$, formed by re-watermarking the input $X_w$, together with the asserted identity $\mathbb{E}\{Y_w|X_w\}=Y$ that converts it into a usable training signal. The architecture's second engine is the Sparse Transformer U-Net, whose top-$k$ sparse self-attention retains only the largest entries of the transposed attention map so that localized watermark patterns can be captured without quadratic spatial cost. The shared encoder and gate mechanism then coordinate the two decoders: the gate uses the watermark-removal features to modulate and damp the denoising features before fusion.

What would settle it

Take one watermarked image $X_w$ whose clean original $Y$ and watermark $W$ are known; generate many random targets $Y_w$ by the paper's recipe and average them. If the average is not $Y$ but instead equals $\alpha W+(1-\alpha)X_w$ averaged, differing from $Y$ whenever the watermark mean is nonzero, the unbiasedness claim is disproved. A second check is to train SSH-Net on synthetic paired data where $Y$ is known and compare its predictions against a model trained directly on the true $Y$ targets.

Watch

Extended reading notes

Core claim

The central claim is that the self-supervised objective in Eq. (4), which minimizes loss between the network output and a synthesized, still-watermarked target $Y_w$, converges to the same solution as supervised training against the true clean image $Y$. The paper argues from the composition model $X_w(p)=\alpha(p)W(p)+(1-\alpha(p))Y(p)$ that randomly adding extra watermarks to $X_w$ produces targets drawn from the same corruption distribution, giving $\mathbb{E}\{Y_w|X_w\}=Y$; this unbiasedness is what lets the synthesized targets stand in for ground truth. Around that idea, SSH-Net decomposes the task into a lightweight CNN noise-removal decoder, a Transformer-based decoder that removes watermarks and noise together using sparse self-attention, a shared encoder, and a gated feature-fusion unit, and reports that this design beats previous self-supervised and supervised baselines on PSNR, SSIM, and LPIPS at several noise levels and watermark opacities.

Load-bearing premise

The load-bearing premise is that a target made by randomly adding more watermarks to the already-watermarked input is, on average, exactly the clean image; if that expectation is wrong, the whole self-supervised training signal is wrong.

Editorial extensions

If this is right

Training watermark-removal models no longer requires paired clean images; any watermarked collection, including images scraped from the web, can supply its own training targets.
The dual-task split means the simpler denoising path can be made lightweight while the harder watermark path carries the transformer cost, a design choice that transfers to other degradation pairs.
Because the sparse attention keeps only top-$k$ attention entries, computation stays low for high-resolution inputs, making the approach more practical for real deployment.
The reported results imply that a single model can handle blind transparency and blind noise ranges simultaneously, removing the need to know the watermark strength or noise level at test time.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the unbiasedness equality fails for real watermarks, which typically have nonzero mean and spatially structured transmittance, the method likely learns a watermark-to-watermark mapping rather than true removal; a direct check would compare predictions against known ground truth on synthetic data.
The same re-corrupt-the-input target recipe could be tried on other structured artifacts with a known blend model, such as subtitles, timestamps, logos, or semi-transparent overlays, where supervised pairs are equally scarce.
The gating behavior the paper visualizes could be turned into a per-pixel confidence map that flags where the model is uncertain, potentially useful for downstream tasks like detecting whether a watermark was present.
The top-$k$ sparsity choice assumes watermark attention is localized; measuring whether attention actually concentrates inside watermark regions would tell whether the inductive bias matches watermark structure.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

I gave this a careful read, and the bottom line is: the architecture is a plausible variant of PSLNet, the experiments are extensive, but the self-supervised justification collapses under the paper's own equations.

What's new: a hybrid dual-decoder where a light CNN branch handles denoising and a heavier sparse-Transformer U-Net tackles watermark plus noise, with a shared encoder and a gate fusion. The sparse top-k attention is known from the Transformer literature, but the combination is a sensible engineering choice. The ablation study is well done, and the authors deserve credit for including complexity and runtime tables and for admitting that runtime is higher than PSLNet's even with lower FLOPs.

Where it falls apart: Sec. 3.1. The paper says that synthesizing Y_w by randomly adding more watermarks to X_w gives E[Y_w|X_w]=Y, so Y_w is an unbiased estimator of the clean image. That is wrong. With the blend model of Eq. (1), X_w = alpha W + (1-alpha)Y, and Y_w = beta W' + (1-beta)X_w for an independent watermark W', we get E[Y_w|X_w] = beta E[W'] + (1-beta)X_w, which is a re-watermarked image, not the clean Y. So the training target in Eq. (4) is a doubly watermarked image, and the impressive numbers, especially 49.38 dB at sigma=0, cannot be explained by the stated protocol. Either the implementation uses a different target construction than described, or there is undisclosed clean-image supervision. This is the central claim of the paper, so the rejection is warranted.

Smaller issues: no error bars, code not available despite the promise, and the efficiency selling point is undercut by the 4x slower runtime and 2.3x more parameters than PSLNet. At alpha=1.0 in Table 2, SSH-Net is actually worse than PSLNet.

Who this is for: someone working on self-supervised restoration would want to see whether the architecture can be re-derived with a valid target. It deserves a serious referee because the experimental evidence is detailed and the architecture, if paired with a correct training scheme, could be useful. But as submitted, the internal inconsistency is fatal to the main claim. I'd send it to review, but the likely outcome is major revision at best.

Referee Report

3 major / 7 minor

Summary. The paper proposes SSH-Net, a self-supervised hybrid network for removing both visible watermarks and Gaussian noise from images. The claimed novelty is a training scheme that synthesizes a reference watermark-free image Y_w by randomly adding additional watermarks to the already-watermarked image X_w; Sec. 3.1 asserts that this construction makes Y_w an unbiased estimator of the clean image Y, i.e., E[Y_w|X_w] = Y, enabling the empirical risk objective in Eq. (4) to be trained without paired clean data. Architecturally, the method combines a shared CNN encoder, a lightweight CNN denoising decoder (NRD), a Transformer-based watermark-and-noise decoder (WNRD) with sparse self-attention, and a gated feature fusion unit. Experiments on PASCAL VOC benchmarks across multiple noise levels and watermark transparencies report consistent improvements over PSLNet and other baselines, and ablations show the contribution of each architectural component.

Significance. If the self-supervised target construction were valid, the paper would make a useful contribution: the hybrid dual-decoder design, sparse attention, and the broad benchmark comparison are sensible and potentially impactful for practical watermark removal without paired data. However, the central statistical claim is internally inconsistent with the paper's own equations, and the empirical gains are not explained by the described training protocol. The manuscript also provides no code, no error bars, and no machine-checked proofs, so the strength of the empirical claims cannot be independently assessed. The significance rests entirely on the unbiasedness assertion, which fails, and therefore the contribution, as currently stated, cannot be credited.

major comments (3)

[Sec. 3.1, Eqs. (1)-(4)]
[Sec. 3.5, Eqs. (17)-(21)]
[Sec. 4.4, Table 10]

minor comments (7)

[Sec. 1] There is a duplicated sentence fragment: 'by learning directly from the corrupted image itself. by learning directly from the corrupted image itself.' should be reduced to one occurrence.
[Sec. 3.3] The word 'nstead' should be 'Instead' in the sentence describing the NAFBlock.
[Sec. 3.4] The module name 'WNRN' appears in the text and should be 'WNRD' for consistency with the rest of the paper.
[Figure 1 caption] The caption contains the typo 'groud-truth'; it should be 'ground-truth'.
[Eq. (15)] The notation 'top-k (row j)' is unclear because the attention map M has dimensions d_l × d_l and the softmax is applied over one dimension; please specify the exact axis along which the top-k selection is performed and whether the mask is applied before or after the softmax.
[Sec. 3.2 and Eq. (19)] The output of the watermark-and-noise decoder is denoted Y_wn in Sec. 3.2 but Y_wm in Eq. (19); please use a single consistent notation.
[Sec. 4.1] The test-set description says 21 PASCAL VOC images are processed with a random watermark and a fixed noise level, yet the total is stated as 252 test images per scenario; the arithmetic (21 × 4 × 3 = 252) is not explained and should be clarified.

Circularity Check

1 steps flagged · score 8.0 of 10

Sec. 3.1 asserts E{Y_w|X_w}=Y for a target obtained by re-watermarking X_w, yet the paper's own blend equation makes E{Y_w|X_w} a re-watermarked input; the central self-supervised derivation is circular.

self definitional [Sec. 3.1, Eqs. (1), (3), (4); target synthesis; losses in Sec. 3.5, Eqs. (18)-(21)]
"Specifically, the synthesized reference image Y_w is generated by randomly adding additional watermarks into the watermarked image X_w. This ensures that both the input image and the synthesized output image are drawn from the same watermark corruption distribution, approximating the condition E{Y_w|X_w} = Y. In other words, Y_w serves as an unbiased estimator of the clean image Y, conditional on the watermarked image X_w."

Under Eq. (1), X_w = αW + (1−α)Y. If Y_w is another watermark blended into X_w, Y_w = βW′ + (1−β)X_w, so E[Y_w|X_w] = βE[W′] + (1−β)X_w, a re-watermarked input, not Y. The asserted unbiasedness would require X_w to already be the clean image up to an affine map, i.e., exactly the inversion to be learned. Thus moving from Eq. (3) to Eq. (4) assumes the conclusion: Y_w is defined from X_w and then declared to represent Y on average. For the L2 loss in Eq. (4), the minimizer is E[Y_w|X_w] ≈ X_w, so the protocol trains toward a doubly watermarked image; reported clean-image gains do not follow from the stated derivation.

full rationale

The central self-supervised claim collapses at the construction of Y_w. The paper defines the synthetic reference by randomly adding further watermarks to X_w and then asserts E{Y_w|X_w}=Y. By the paper's own blend model (Eq. (1)), that conditional expectation is a re-watermarked version of the input rather than the clean image; the unbiasedness condition is not derived but is exactly the property to be established. All subsequent losses (Eqs. (18)-(21)) compare outputs against Y_w, so the same circular target propagates through every supervision term. This is not a self-citation or uniqueness-import problem: PSLNet and Noise2Noise are external prior work, and the architecture ablations are empirically self-contained. The flaw is self-definitional: the 'synthesized reference watermark-free image' is, by construction, a doubly watermarked image, so Eq. (4) trains toward the input's watermarked version. The stark sigma=0 result (49.38 dB) would require an undocumented target-generation protocol or clean-image supervision; it cannot be derived from the stated unbiasedness argument. Score 8 reflects that the paper's headline claim reduces by definition to a target that is not the clean image.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The ledger is heavy because the paper's self-supervised 'derivation' depends on an unproven unbiasedness assumption and several hand-picked hyperparameters. No new entities are posited.

free parameters (6)

loss weight alpha = 0.024
Balances structural L1 and perceptual VGG losses (Sec. 4.2); tuned by hand, not derived.
top-k rates for sparse attention = (1/2, 2/3, 3/4, 4/5)
Selected as sequence of sparsity levels for the four attention heads (Sec. 4.2); hand-chosen.
per-level block counts (L1-L5) = (2,4,4,6,6)
Depth of encoder and decoders (Sec. 4.2); architecture hyperparameter.
number of attention heads = (4,8,8,8,4)
Sparse Transformer U-Net heads (Sec. 4.2); architecture hyperparameter.
FFN expansion ratio = 2.66
Feed-forward expansion in ST blocks (Sec. 4.2); hand-chosen.
channel width = 48
Feature channels in shared encoder and decoders (Sec. 4.2); architecture hyperparameter.

assumptions (4)

ad hoc to paper The synthesized target Y_w is an unbiased estimator of the clean image Y: E{Y_w|X_w}=Y
Asserted in Sec. 3.1 to justify Eq. (4); contradicted by the paper's own blend model Eq. (1), since Y_w is a re-watermarking of X_w, so E[Y_w|X_w]=(1-alpha)X_w + alpha E[W].
domain assumption Additive Gaussian noise K is zero-mean and independent of image content
Used in X_wn = X_w + K and in the Noise2Noise-style argument; reasonable for the synthetic benchmark but not tested on real noise distributions.
domain assumption Watermark blend model X_w(p) = alpha(p)W(p)+(1-alpha(p))Y(p) holds
Assumed in Eq. (1); all training and test data are synthesized under this model, so reported results may not transfer to real watermarks with different alpha maps or blending operations.
ad hoc to paper Minimizing a mixed L1 and perceptual loss against an unbiased target recovers the clean image
Noise2Noise unbiasedness is proved for L2, not L1; the paper uses L1 plus VGG loss but does not address the conditional-median problem, so the theoretical guarantee is incomplete.

how reviews work

0 comments

Cite this review

Pith. "Pith review of SSH-Net: A Self-Supervised and Hybrid Network for Noisy Image Watermark Removal." pith.science (2026). https://pith.science/paper/GEJJN7D6

@misc{pith2026250505088,
  author       = {Pith},
  title        = {Pith review of: SSH-Net: A Self-Supervised and Hybrid Network for Noisy Image Watermark Removal},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GEJJN7D6}},
  note         = {Machine review of arXiv:2505.05088}
}

read the original abstract

Visible watermark removal is challenging due to its inherent complexities and the noise carried within images. Existing methods primarily rely on supervised learning approaches that require paired datasets of watermarked and watermark-free images, which are often impractical to obtain in real-world scenarios. To address this challenge, we propose SSH-Net, a Self-Supervised and Hybrid Network specifically designed for noisy image watermark removal. SSH-Net synthesizes reference watermark-free images using the watermark distribution in a self-supervised manner and adopts a dual-network design to address the task. The upper network, focused on the simpler task of noise removal, employs a lightweight CNN-based architecture, while the lower network, designed to handle the more complex task of simultaneously removing watermarks and noise, incorporates Transformer blocks to model long-range dependencies and capture intricate image features. To enhance the model's effectiveness, a shared CNN-based feature encoder is introduced before dual networks to extract common features that both networks can leverage. Our code will be available at https://github.com/wenyang001/SSH-Net.

Figures

Figures reproduced from arXiv: 2505.05088 by the authors.

**Figure 1.** The overall architecture of the proposed Self-Supervised and Hybrid Network (SSH-Net), which mainly contains a shared encoder, a noise removal decoder (NRD), a watermark and noise removal decoder (WNRD), and a feature fusion unit (FFU). SSH-Net leveraged the supervised and self-supervised technique to generate groud-truth samples (𝑋w , 𝑌w ) to optimize the model. model to approximate the real ground-truth Y, even wi… view at source ↗

**Figure 2.** The overall architecture of the proposed Sparse Transformer U-Net, which follows a 3-level Transformer U-Net design. Each level comprises multiple Sparse Transformer Blocks (STBs) to process features at different scales. Each STB includes a Sparse Self-Attention (SSA) mechanism and a Fully-Connected Network (FFN). 3.3. Noise Removal Decoder The noise removal decoder (NRD), with a structure similar to the SE, employ… view at source ↗

**Figure 3.** Results comparison trained under a specific transparency and a specific noise condition (𝛿 = 25 and transparency = 0.3). (a) Ground Truth (b) 20.19 dB (c) 28.48 dB (d) 29.89 dB (e) 29.68 dB (f) 31.39 dB (g) 32.15 dB (h) 32.26 dB. (a) Original Input (b) Corrupted Input (c) DnCNN (d) FFDNet (e) IRCNN (f) DRDNet (g) PSLNet (h) SSH-Net (Ours) [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Results comparison trained under a specific noise and a specific transparency condition (𝛿 = 15 and transparency = 0.3). (a) Ground Truth (b) 24.04 dB (c) 27.44 dB (d) 27.19 dB (e) 30.16 dB (f) 29.23 dB (g) 31.54 dB (h) 31.95 dB. randomly between 0.3 and 1.0. For testi…

**Figure 5.** Figure 5: Visualizations of the gating mechanism in the proposed Feature Fusion Unit (FFU). better watermark removal performance across various noise and transparency levels. Furthermore, SSH-Net achieves fewer FLOPs than DnCNN, DRDNet, and PSLNet, owing to its hybrid structure …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 31 canonical work pages

[1]

Protecting publicly-available images with an invisible image watermark

Gordon W Braudaway. Protecting publicly-available images with an invisible image watermark. In Proceedings of international conference on image processing, volume 1, pages 524--527. IEEE, 1997

work page 1997
[2]

Tsrformer: Transformer based two-stage refinement for single image shadow removal

Hua-En Chang, Chia-Hsuan Hsieh, Hao-Hsiang Yang, I Chen, Yi-Chung Chen, Yuan-Chun Chiang, Zhi-Kai Huang, Wei-Ting Chen, Sy-Yen Kuo, et al. Tsrformer: Transformer based two-stage refinement for single image shadow removal. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1436--1446, 2023

work page 2023
[3]

Simple baselines for image restoration

Liangyu Chen, Xiaojie Chu, Xiangyu Zhang, and Jian Sun. Simple baselines for image restoration. In European conference on computer vision, pages 17--33. Springer, 2022

work page 2022
[4]

Learning a sparse transformer network for effective image deraining

Xiang Chen, Hao Li, Mingqiang Li, and Jinshan Pan. Learning a sparse transformer network for effective image deraining. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5896--5905, 2023

work page 2023
[5]

Leveraging unlabeled data for watermark removal of deep neural networks

Xinyun Chen, Wenxiao Wang, Yiming Ding, Chris Bender, Ruoxi Jia, Bo Li, and Dawn Song. Leveraging unlabeled data for watermark removal of deep neural networks. In ICML workshop on Security and Privacy of Machine Learning, pages 1--6, 2019

work page 2019
[6]

Large-scale visible watermark detection and removal with deep convolutional networks

Danni Cheng, Xiang Li, Wei-Hong Li, Chan Lu, Fake Li, Hua Zhao, and Wei-Shi Zheng. Large-scale visible watermark detection and removal with deep convolutional networks. In Pattern Recognition and Computer Vision: First Chinese Conference, PRCV 2018, Guangzhou, China, November 23-26, 2018, Proceedings, Part III 1, pages 27--40. Springer, 2018

work page 2018
[7]

On the effectiveness of visible watermarks

Tali Dekel, Michael Rubinstein, Ce Liu, and William T Freeman. On the effectiveness of visible watermarks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 2146--2154, 2017

work page 2017
[8]

DRD-Net: Detail-recovery Image Deraining via Context Aggregation Networks

Sen Deng, Mingqiang Wei, Jun Wang, Luming Liang, Haoran Xie, and Meng Wang. Drd-net: Detail-recovery image deraining via context aggregation networks. arXiv preprint arXiv:1908.10267, 2019

work page Pith review arXiv 1908

Show all 42 references

[9]

An image is worth 16x16 words: Transformers for image recognition at scale

Alexey Dosovitskiy. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2020

2010 arXiv
[10]

The pascal visual object classes challenge: A retrospective

Mark Everingham, SM Ali Eslami, Luc Van Gool, Christopher KI Williams, John Winn, and Andrew Zisserman. The pascal visual object classes challenge: A retrospective. International journal of computer vision, 111: 0 98--136, 2015

2015
[11]

Real-time attacks on robust watermarking tools in the wild by cnn

Linfeng Geng, Weiming Zhang, Haozhe Chen, Han Fang, and Nenghai Yu. Real-time attacks on robust watermarking tools in the wild by cnn. Journal of Real-Time Image Processing, 17: 0 631--641, 2020

2020
[12]

Image quality metrics: Psnr vs

Alain Hore and Djemel Ziou. Image quality metrics: Psnr vs. ssim. In 2010 20th international conference on pattern recognition, pages 2366--2369. IEEE, 2010

2010
[13]

New watermark-removal method based on eigen-image energy

T-C Hsu, W-S Hsieh, John Y Chiang, and T-S Su. New watermark-removal method based on eigen-image energy. IET Information Security, 5 0 (1): 0 43--50, 2011

2011
[14]

An algorithm for removable visible watermarking

Yongjian Hu, Sam Kwong, and Jiwu Huang. An algorithm for removable visible watermarking. IEEE Transactions on Circuits and Systems for Video Technology, 16 0 (1): 0 129--133, 2005

2005
[15]

Attacking visible watermarking schemes

Chun-Hsiang Huang and Ja-Ling Wu. Attacking visible watermarking schemes. IEEE transactions on multimedia, 6 0 (1): 0 16--30, 2004

2004
[16]

Noise2void-learning denoising from single noisy images

Alexander Krull, Tim-Oliver Buchholz, and Florian Jug. Noise2void-learning denoising from single noisy images. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 2129--2137, 2019

2019
[17]

Noise2noise: Learning image restoration without clean data

Jaakko Lehtinen, Jacob Munkberg, Jon Hasselgren, Samuli Laine, Tero Karras, Miika Aittala, and Timo Aila. Noise2noise: Learning image restoration without clean data. In International Conference on Machine Learning, pages 2965--2974. PMLR, 2018

2018
[18]

Towards photo-realistic visible watermark removal with conditional generative adversarial networks

Xiang Li, Chan Lu, Danni Cheng, Wei-Hong Li, Mei Cao, Bo Liu, Jiechao Ma, and Wei-Shi Zheng. Towards photo-realistic visible watermark removal with conditional generative adversarial networks. In Image and Graphics: 10th International Conference, ICIG 2019, Beijing, China, Aug...

2019
[19]

Swinir: Image restoration using swin transformer

Jingyun Liang, Jiezhang Cao, Guolei Sun, Kai Zhang, Luc Van Gool, and Radu Timofte. Swinir: Image restoration using swin transformer. In Proceedings of the IEEE/CVF international conference on computer vision, pages 1833--1844, 2021

2021
[20]

Self-supervised learning and adaptation for single image dehazing

Yudong Liang, Bin Wang, Wangmeng Zuo, Jiaying Liu, and Wenqi Ren. Self-supervised learning and adaptation for single image dehazing. In IJCAI, pages 1137--1143, 2022

2022
[21]

Bitstream-corrupted jpeg images are restorable: Two-stage compensation and alignment framework for image restoration

Wenyang Liu, Yi Wang, Kim-Hui Yap, and Lap-Pui Chau. Bitstream-corrupted jpeg images are restorable: Two-stage compensation and alignment framework for image restoration. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 9979--9988, 2023

2023
[22]

Bytenet: Rethinking multimedia file fragment classification through visual perspectives

Wenyang Liu, Kejun Wu, Tianyi Liu, Yi Wang, Kim-Hui Yap, and Lap-Pui Chau. Bytenet: Rethinking multimedia file fragment classification through visual perspectives. IEEE Transactions on Multimedia, pages 1--14, 2024. doi:10.1109/TMM.2024.3521830

2024
[23]

Swin transformer: Hierarchical vision transformer using shifted windows

Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF international conference on computer vision, pages 10012--10022, 2021

2021
[24]

A novel image recovery algorithm for visible watermarked images

Soo-Chang Pei and Yi-Chong Zeng. A novel image recovery algorithm for visible watermarked images. IEEE Transactions on information forensics and security, 1 0 (4): 0 543--550, 2006

2006
[25]

A survey of digital watermarking techniques, applications and attacks

Prabhishek Singh and Ramneet Singh Chadha. A survey of digital watermarking techniques, applications and attacks. International Journal of Engineering and Innovative Technology (IJEIT), 2 0 (9): 0 165--175, 2013

2013
[26]

Vision transformers for single image dehazing

Yuda Song, Zhuqing He, Hui Qian, and Xin Du. Vision transformers for single image dehazing. IEEE Transactions on Image Processing, 32: 0 1927--1941, 2023

1927
[27]

A self-supervised network for image denoising and watermark removal

Chunwei Tian, Jingyu Xiao, Bob Zhang, Wangmeng Zuo, Yudong Zhang, and Chia-Wen Lin. A self-supervised network for image denoising and watermark removal. Neural Networks, 174: 0 106218, 2024 a

2024
[28]

Perceptive self-supervised learning network for noisy image watermark removal

Chunwei Tian, Menghua Zheng, Bo Li, Yanning Zhang, Shichao Zhang, and David Zhang. Perceptive self-supervised learning network for noisy image watermark removal. IEEE Transactions on Circuits and Systems for Video Technology, 2024 b

2024
[29]

Transweather: Transformer-based restoration of images degraded by adverse weather conditions

Jeya Maria Jose Valanarasu, Rajeev Yasarla, and Vishal M Patel. Transweather: Transformer-based restoration of images degraded by adverse weather conditions. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2353--2363, 2022

2022
[30]

Crformer: A cross-region transformer for shadow removal

Jin Wan, Hui Yin, Zhenyao Wu, Xinyi Wu, Zhihao Liu, and Song Wang. Crformer: A cross-region transformer for shadow removal. Image and Vision Computing, 151: 0 105273, 2024

2024
[31]

Fastderainnet: A deep learning algorithm for single image deraining

Xiuwen Wang, Zhiwei Li, Hongtao Shan, Zhiyuan Tian, Yuanhong Ren, and Wuneng Zhou. Fastderainnet: A deep learning algorithm for single image deraining. IEEE Access, 8: 0 127622--127630, 2020

2020
[32]

Image quality assessment: from error visibility to structural similarity

Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Simoncelli. Image quality assessment: from error visibility to structural similarity. IEEE transactions on image processing, 13 0 (4): 0 600--612, 2004

2004
[33]

Image de-raining transformer

Jie Xiao, Xueyang Fu, Aiping Liu, Feng Wu, and Zheng-Jun Zha. Image de-raining transformer. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45 0 (11): 0 12978--12995, 2022

2022
[34]

An automatic visible watermark removal technique using image inpainting algorithms

Chaoran Xu, Yao Lu, and Yuanpin Zhou. An automatic visible watermark removal technique using image inpainting algorithms. In 2017 4th International Conference on Systems and Informatics (ICSAI), pages 1152--1157. IEEE, 2017

2017
[35]

Metaformer is actually what you need for vision

Weihao Yu, Mi Luo, Pan Zhou, Chenyang Si, Yichen Zhou, Xinchao Wang, Jiashi Feng, and Shuicheng Yan. Metaformer is actually what you need for vision. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10819--10829, 2022

2022
[36]

Restormer: Efficient transformer for high-resolution image restoration

Syed Waqas Zamir, Aditya Arora, Salman Khan, Munawar Hayat, Fahad Shahbaz Khan, and Ming-Hsuan Yang. Restormer: Efficient transformer for high-resolution image restoration. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5728--5739, 2022

2022
[37]

Beyond a gaussian denoiser: Residual learning of deep cnn for image denoising

Kai Zhang, Wangmeng Zuo, Yunjin Chen, Deyu Meng, and Lei Zhang. Beyond a gaussian denoiser: Residual learning of deep cnn for image denoising. IEEE transactions on image processing, 26 0 (7): 0 3142--3155, 2017 a

2017
[38]

Learning deep cnn denoiser prior for image restoration

Kai Zhang, Wangmeng Zuo, Shuhang Gu, and Lei Zhang. Learning deep cnn denoiser prior for image restoration. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 3929--3938, 2017 b

2017
[39]

Ffdnet: Toward a fast and flexible solution for cnn-based image denoising

Kai Zhang, Wangmeng Zuo, and Lei Zhang. Ffdnet: Toward a fast and flexible solution for cnn-based image denoising. IEEE Transactions on Image Processing, 27 0 (9): 0 4608--4622, 2018 a

2018
[40]

The unreasonable effectiveness of deep features as a perceptual metric

Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shechtman, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. In CVPR, 2018 b

2018
[41]

Object detection with deep learning: A review

Zhong-Qiu Zhao, Peng Zheng, Shou-tao Xu, and Xindong Wu. Object detection with deep learning: A review. IEEE transactions on neural networks and learning systems, 30 0 (11): 0 3212--3232, 2019

2019
[42]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [1]

Protecting publicly-available images with an invisible image watermark

Gordon W Braudaway. Protecting publicly-available images with an invisible image watermark. In Proceedings of international conference on image processing, volume 1, pages 524--527. IEEE, 1997

work page 1997

[2] [2]

Tsrformer: Transformer based two-stage refinement for single image shadow removal

Hua-En Chang, Chia-Hsuan Hsieh, Hao-Hsiang Yang, I Chen, Yi-Chung Chen, Yuan-Chun Chiang, Zhi-Kai Huang, Wei-Ting Chen, Sy-Yen Kuo, et al. Tsrformer: Transformer based two-stage refinement for single image shadow removal. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1436--1446, 2023

work page 2023

[3] [3]

Simple baselines for image restoration

Liangyu Chen, Xiaojie Chu, Xiangyu Zhang, and Jian Sun. Simple baselines for image restoration. In European conference on computer vision, pages 17--33. Springer, 2022

work page 2022

[4] [4]

Learning a sparse transformer network for effective image deraining

Xiang Chen, Hao Li, Mingqiang Li, and Jinshan Pan. Learning a sparse transformer network for effective image deraining. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5896--5905, 2023

work page 2023

[5] [5]

Leveraging unlabeled data for watermark removal of deep neural networks

Xinyun Chen, Wenxiao Wang, Yiming Ding, Chris Bender, Ruoxi Jia, Bo Li, and Dawn Song. Leveraging unlabeled data for watermark removal of deep neural networks. In ICML workshop on Security and Privacy of Machine Learning, pages 1--6, 2019

work page 2019

[6] [6]

Large-scale visible watermark detection and removal with deep convolutional networks

Danni Cheng, Xiang Li, Wei-Hong Li, Chan Lu, Fake Li, Hua Zhao, and Wei-Shi Zheng. Large-scale visible watermark detection and removal with deep convolutional networks. In Pattern Recognition and Computer Vision: First Chinese Conference, PRCV 2018, Guangzhou, China, November 23-26, 2018, Proceedings, Part III 1, pages 27--40. Springer, 2018

work page 2018

[7] [7]

On the effectiveness of visible watermarks

Tali Dekel, Michael Rubinstein, Ce Liu, and William T Freeman. On the effectiveness of visible watermarks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 2146--2154, 2017

work page 2017

[8] [8]

DRD-Net: Detail-recovery Image Deraining via Context Aggregation Networks

Sen Deng, Mingqiang Wei, Jun Wang, Luming Liang, Haoran Xie, and Meng Wang. Drd-net: Detail-recovery image deraining via context aggregation networks. arXiv preprint arXiv:1908.10267, 2019

work page Pith review arXiv 1908