Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

Unpaired Deblurring via Decoupled Diffusion Model

T0 review · 3 major / 6 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a diffusion-based deblurring model can generalize to unseen blur domains by learning structural and blur features separately through three jointly optimized tasks.

desk verdict Novel decoupled training for unpaired deblurring, but the SOTA claim is undercut by the paper's own reference-based numbers. read the letter →

arxiv 2502.01522 v2 pith:GRWXTRLA submitted 2025-02-03 cs.CV

Junhao Cheng , Wei-Ting Chen , Xi Lu , Ming-Hsuan Yang This is my paper · ORCID
classification cs.CV
keywords unpairedimagedeblurringdiffusionmodelsblurpatterndecouplingQ-Formerunsupervisedtransferreal-worldgeneralizationtounknowndomainsgenerativeprior
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

The paper proposes UID-Diff, a generative-diffusion model for deblurring images from domains where no blurry-sharp pairs are available. Its central claim is that separating structural content from blur patterns, rather than learning a direct blurry-to-sharp mapping, makes the model generalize to unknown blur types. Two Q-Former extractors are trained together on a supervised deblurring task, an unsupervised blur-transfer task, and a reconstruction task, so that structure features and blur features become complementary. On real-world benchmarks the authors report that UID-Diff removes blur and preserves structure better than existing methods, including other diffusion-based and unpaired approaches.

What carries the argument

Two Q-Former extractors separate the image into structure features fs and blur features fb; a ControlNet-style adapter injects fs into the frozen Stable Diffusion U-Net, while fb enters through newly added cross-attention key and value projections. Three losses drive the decoupling: a supervised diffusion denoising loss for deblurring, an unsupervised blur-transfer loss that re-renders a target-domain image with a blur pattern taken from another image, and a reconstruction loss that combines both features to reproduce the original blurry image. A refined VAE replaces the stock Stable Diffusion VAE to reduce detail drift during decoding.

What would settle it

A classifier trained on the blur-extractor features that can reliably identify which sharp image a blurred image came from would falsify the claim that Qb isolates blur independently of structure; the paper does not report such a content-leakage test.

Watch

Extended reading notes

Core claim

UID-Diff's core discovery is that blur can be treated as a separate, transferable condition for a frozen text-to-image diffusion model, distinct from the structural features that deblurring must preserve. The paper argues that prior diffusion-based deblurring adapters overfit synthetic blur because they condition on entangled blurry images; by contrast, a structure extractor Qs learned from synthetic pairs and a blur-pattern extractor Qb learned from unpaired target-domain images can be made complementary through a reconstruction loss, so that at inference only Qs and the adapter are needed to restore a sharp image. The claim is that this decoupled learning directly improves generalization to unseen blur patterns while keeping the generative detail of Stable Diffusion.

Load-bearing premise

The whole method depends on the blur-transfer and reconstruction tasks actually forcing the blur extractor to capture only blur and the structure extractor to ignore blur, with only qualitative visual evidence supporting this separation.

Editorial extensions

If this is right

  • The method can adapt to a new blurry camera or environment using only unpaired blurred images, without collecting sharp counterparts.
  • Because the blur extractor is trained unsupervised, the same pipeline can be repurposed for other degradations, such as noise, haze, or compression artifacts, treated as transferable style conditions.
  • The reconstruction task provides a consistency mechanism that prevents structure and blur features from collapsing into each other, yielding sharper text and edges in restored images.
  • Diffusion-based deblurring becomes viable in settings where paired data is scarce, with the generative model supplying fine detail that supervised non-generative methods miss.

Reading between the lines

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

  • A quantitative test of whether Qb's output is free of structural content, for instance decoding the blur feature alone or measuring perceptual similarity between blur-transfer outputs and their structure sources, would directly settle whether true complementarity is achieved rather than a learned shortcut.
  • The decoupling idea suggests that a single structure extractor could serve multiple degradation-specific blur extractors, turning the framework into a modular restoration system; the paper does not explore this.
  • Because the reported metrics are no-reference or reference-based on existing benchmarks, a natural extension is measuring whether the improved generalization persists under a held-out camera or kernel distribution shift with a perceptual fidelity metric aligned to human judgment.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The manuscript proposes UID-Diff, a Stable-Diffusion-based image deblurring method that aims to handle unknown blur domains without paired data. The method uses two Q-Former extractors, one intended to capture structural features (Qs) and one intended to capture blur patterns (Qb), trained jointly with three objectives: a supervised deblurring loss on synthetic blurry-sharp pairs, an unsupervised blur-transfer loss using unpaired target-domain blurred images, and a reconstruction loss intended to enforce complementarity of the two extracted representations. Experiments are reported on GoPro, REDS, and RealBlur-J/R, with comparisons against supervised and unsupervised deblurring methods, ablations, and qualitative visualizations. The code is released.

Significance. The motivation is timely and practically relevant: real paired blurry-sharp data are scarce, and using unpaired target-domain blurred images to improve a diffusion-based deblurring adapter could reduce overfitting to synthetic blur. The paper has several strengths: it presents a clearly described training pipeline, includes extensive ablations, releases code, and the three-task formulation is a reasonable attempt to separate structure from blur. However, the central quantitative claim that UID-Diff outperforms existing state-of-the-art methods in blur removal and structural preservation is not supported by the paper's own reference-based evaluation, and the decoupling mechanism, which is the core novelty, is validated only qualitatively. The paper would need substantial revisions to align its claims with its evidence.

major comments (3)
  1. [Section 4.3 and Appendix B, Table 5] The abstract and Section 5 state that UID-Diff outperforms existing SOTA methods in blur removal and structural preservation, but the main quantitative support (Table 1) is exclusively no-reference IQA metrics (MANIQA, LIQE, MUSIQ, CLIP-IQA). The paper's own reference-based comparison in Appendix Table 5 contradicts the headline: on GoPro, AdaRevD achieves PSNR/SSIM 26.93/0.7492 versus UID-Diff's 25.08/0.7403; on REDS, Blur2Blur achieves 28.94/0.8522 and FFTformer 27.53/0.8320 versus 25.48/0.7822; on RealBlur-R, FFTformer and AdaRevD achieve 26.94/0.7542 and 27.85/0.8034 versus 22.47/0.7384; FID also favors AdaRevD on GoPro (31.92 versus 49.68). Blur removal and structural preservation are fidelity-centric notions, and no-reference aesthetics metrics can reward synthesized texture rather than faithful restoration. This is an internal inconsistency between the abstract and the appendix, not merely a disagreement with external consensus. The authors should either revise the SOTA claim to reflect the metric families in which UID-Diff actually leads, or provide reference-based evidence that supports the original claim.
  2. [Sections 3.5 and 3.6, Figures 9-10] The core technical assumption is that Qb learns blur patterns independently of image structure and that Qs learns structure independently of blur, with complementarity enforced by the reconstruction loss in Eq. (7). The only supporting evidence is the qualitative blur-transfer visualization in Appendix C, Figures 9-10. No quantitative measurement is given to show that the transferred blur is free of the source image's structure, that Qs is invariant to blur type, or that the reconstruction task genuinely enforces complementarity rather than merely acting as an image autoencoding loss. Without such evidence, the method's claimed generalization mechanism is not established, and the method could reduce to a supervised deblurring adapter with auxiliary losses. Please add quantitative decoupling tests, for example measuring structure leakage from the source sharp image into the blur-transferred output (using LPIPS or a structure classifier), evaluating Qs features under different blur kernels, or ablating Qb at inference with reference-based metrics.
  3. [Section 4.2 and Table 2] The 'unknown domain' evaluation protocol needs clarification. The synthetic blur data B' is generated from the sharp subset S of each target dataset, and the test set is drawn from the same dataset. This means the training data and test data share content statistics, which could make the generalization results look better than they would for a truly unseen content domain. The authors state they follow [2], but they should explicitly describe the split, confirm that no content overlap exists between the training split and the test split, and discuss how this affects the conclusion that UID-Diff generalizes to unknown blur domains.
minor comments (6)
  1. [Section 3.5 heading] The heading 'Blur-tansfer' appears to be a typo and should read 'Blur-transfer'.
  2. [Table 1] The metric name 'MSUIQ' in Table 1 should be 'MUSIQ' for consistency with Section 4.1 and the rest of the paper.
  3. [Section 3.7] The phrase 'the newly added projection metrics' should be 'projection matrices', since the text refers to trainable weight matrices in the cross-attention layers.
  4. [Equation (7)] The feature variable 'fc' in Eq. (7) is inconsistent with the notation 'fs' used for structural features in Sections 3.4 and in Eq. (4). Please unify the notation.
  5. [Appendix B] The text says the comparison with SOTA methods on reference-based metrics is 'presented in Table 4', but the SOTA reference-based table is actually Table 5; the table numbering and cross-references should be corrected.
  6. [References] Reference [7] cites an AIRNet paper on self-supervised affine registration for 3D medical images, which is not the AIRNet image-restoration method evaluated in the paper. The citation should be corrected.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: UID-Diff is an empirical systems paper evaluated on external benchmarks, and its self-citations are not load-bearing.

full rationale

This is an empirical systems paper: the central claim of superior blur removal and structural preservation is established by benchmarking against external methods on GoPro, REDS, RealBlur-J, and RealBlur-R, not by deriving a quantity from itself. The training objectives L1-L3 in Eqs. (4), (6), and (7) use external blurry-sharp pairs and unpaired target-domain blur; no fitted parameter is renamed as a prediction, and no equation reduces to another by construction. The decoupling assumption in Sections 3.5-3.6 is only qualitatively validated in Appendix C, but an unsupported assumption is a validation concern, not circularity. The only self-citations touching the architecture are the refined VAE from [70] (same author group) and image-generation works [27,64]; the refined VAE is ablated in Tables 3-4 and is not load-bearing for the central benchmark claim, so it does not make the evaluation circular. The contradiction between the no-reference Table 1 and the reference-based Table 5 concerns the strength of the evidence for the SOTA claim, not a circular derivation.

Assumptions & free parameters 3 free parameters · 4 assumptions · 2 invented entities

The central claim depends on the assumption that two learned latent codes cleanly separate structure and blur. This is the paper's main postulated entity. No formal or independent evidence for the separation is given, and the blur pattern representation is never quantitatively evaluated for independence from content.

free parameters (3)
  • Loss weights alpha, beta, gamma = 1:1:1
    Chosen by hand in Section 3.7; no sensitivity analysis is reported, yet the joint loss balance is central to the training scheme.
  • Classifier-free guidance scale = 7.5
    Chosen for inference in Section 4.2; affects output sharpness and fidelity, no ablation across values.
  • Number of Q-Former query tokens = 16
    Set in Section 4.2 consistent with BLIP-Diffusion; affects capacity of feature extraction.
assumptions (4)
  • domain assumption Blur patterns are domain-specific and can be transferred between images independently of structure.
    Core assumption for the blur-transfer task (Section 3.5); no direct quantitative validation.
  • domain assumption The text prompts 'structure' and 'blur' guide the Q-Formers to extract the intended factors.
    Used in Sections 3.4 and 3.5; not validated by any feature analysis.
  • domain assumption The reconstruction task enforces complementarity between structural and blur features.
    Section 3.6 states this but does not demonstrate that complementarity (rather than simple reconstruction) is achieved.
  • domain assumption Synthetic blur generated from sharp images is representative enough to train the structure extractor.
    Section 4.1 uses synthetic blur; the paper notes that synthetic-only training overfits, yet still relies on it for T1.
invented entities (2)
  • Blur pattern representation f_b
    purpose: A latent vector meant to capture only the domain-specific blur pattern, extracted by Qb and used to condition the diffusion model in blur-transfer.
    The paper postulates this factor can be isolated; evidence is limited to qualitative blur-transfer images (Appendix C) that do not prove the representation is free of content.
  • Structural representation f_s
    purpose: A latent vector meant to carry the underlying sharp structure of a blurry image, extracted by Qs and used as the control condition for deblurring.
    The paper assumes this is invariant to blur; performance under unseen blur is the only indirect evidence, and no direct measurement of f_s content is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Unpaired Deblurring via Decoupled Diffusion Model." pith.science (2026). https://pith.science/paper/GRWXTRLA

@misc{pith2026250201522,
  author       = {Pith},
  title        = {Pith review of: Unpaired Deblurring via Decoupled Diffusion Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GRWXTRLA}},
  note         = {Machine review of arXiv:2502.01522}
}
read the original abstract

Generative diffusion models trained on large-scale datasets have achieved remarkable progress in image synthesis. In favor of their ability to supplement missing details and generate aesthetically pleasing contents, recent works have applied them to image deblurring via training an adapter on blurry-sharp image pairs to provide structural conditions for restoration. However, acquiring substantial amounts of realistic paired data is challenging and costly in real-world scenarios. On the other hand, relying solely on synthetic data often results in overfitting, leading to unsatisfactory performance when confronted with unseen blur patterns. To tackle this issue, we propose UID-Diff, a generative-diffusion-based model designed to enhance deblurring performance on unknown domains by decoupling structural features and blur patterns through joint training on three specially designed tasks. We employ two Q-Formers as structural features and blur patterns extractors separately. The features extracted by them will be used for the supervised deblurring task on synthetic data and the unsupervised blur-transfer task by leveraging unpaired blurred images from the target domain simultaneously. We further introduce a reconstruction task to make the structural features and blur patterns complementary. This blur-decoupled learning process enhances the generalization capabilities of UID-Diff when encountering unknown blur patterns. Experiments on real-world datasets demonstrate that UID-Diff outperforms existing state-of-the-art methods in blur removal and structural preservation in various challenging scenarios.

Figures

Figures reproduced from arXiv: 2502.01522 by the authors.

Figure 1
Figure 1. Comparison of UID-Diff with existing generative-diffusion-based deblurring methods. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Training pipeline of UID-Diff. UID-Diff achieves blur pattern-structure representation [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Inference process of UID-Diff. We preserve the [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (14 more)
Figure 5
Figure 5. Figure 5: Visual comparison of deblurring results. UID-Diff outperforms other SOTA methods in [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Visual comparison on structural reconstruction. UID-Diff outperforms other generative [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Deblurring results of ablation studies. Key regions are marked with red boxes and magnified. [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Architecture of the refined-VAE in UID-Diff. [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 9
Figure 9. Figure 9: Visualization of blur transferring. The blur pattern in the images of the first row is [PITH_FULL_IMAGE:figures/full_fig_p013_9.png]
Figure 10
Figure 10. Figure 10: Visualization of blur transferring. The blur pattern in the images of the first row is [PITH_FULL_IMAGE:figures/full_fig_p013_10.png]
Figure 11
Figure 11. Figure 11: Additional visual comparison on the GoPro dataset. 14 [PITH_FULL_IMAGE:figures/full_fig_p014_11.png]
Figure 12
Figure 12. Figure 12: Additional visual comparison on the GoPro dataset (Continue). 15 [PITH_FULL_IMAGE:figures/full_fig_p015_12.png]
Figure 13
Figure 13. Figure 13: Additional visual comparison on the REDS dataset. 16 [PITH_FULL_IMAGE:figures/full_fig_p016_13.png]
Figure 14
Figure 14. Figure 14: Additional visual comparison on the REDS dataset (Continue). 17 [PITH_FULL_IMAGE:figures/full_fig_p017_14.png]
Figure 15
Figure 15. Figure 15: Additional visual comparison on the RealBlur-J dataset. 18 [PITH_FULL_IMAGE:figures/full_fig_p018_15.png]
Figure 16
Figure 16. Figure 16: Additional visual comparison on the RealBlur-J dataset (Continue). 19 [PITH_FULL_IMAGE:figures/full_fig_p019_16.png]
Figure 17
Figure 17. Figure 17: Additional visual comparison on the RealBlur-R dataset. 20 [PITH_FULL_IMAGE:figures/full_fig_p020_17.png]
Figure 18
Figure 18. Figure 18: Additional visual comparison on the RealBlur-R dataset (Continue). 21 [PITH_FULL_IMAGE:figures/full_fig_p021_18.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Generative Manifold Distillation: Aligning Restoration Trajectories with Natural Image Prior

    cs.CV 2025-12 conditional novelty 6.0 of 10

    Pre-trained restoration models can be adapted to unlabeled out-of-distribution images by distilling quality-gated pseudo-targets from a frozen text-to-image diffusion oracle via mixed-supervision fine-tuning.

Reference graph

Works this paper leans on

82 extracted references · 62 canonical work pages · cited by 1 Pith paper

  1. [47]

    Benchmarking neural network robustness to common corruptions and perturbations

    Dan Hendrycks and Thomas Dietterich. Benchmarking neural network robustness to common corruptions and perturbations. arXiv preprint arXiv:1903.12261, 2019. 3, 5

    arXiv 1903

  2. [71]

    Real-world blur dataset for learning and benchmarking deblurring algorithms

    Jaesung Rim, Haeyun Lee, Jucheol Won, and Sunghyun Cho. Real-world blur dataset for learning and benchmarking deblurring algorithms. In Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XXV 16, pages 184–201. Springer, 2020. 6

    work page 2020

  3. [2]

    Blur2blur: Blur conversion for unsupervised image deblurring on unknown domains

    Bang-Dang Pham, Phong Tran, Anh Tran, Cuong Pham, Rang Nguyen, and Minh Hoai. Blur2blur: Blur conversion for unsupervised image deblurring on unknown domains. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2804–2813, 2024. 1, 2, 3, 4, 6, 7

    2024

  4. [1]

    Neural maximum a posteriori estimation on unpaired data for motion deblurring

    Youjian Zhang, Chaoyue Wang, and Dacheng Tao. Neural maximum a posteriori estimation on unpaired data for motion deblurring. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023. 1, 2, 3

    2023

  5. [3]

    Deblurring by realistic blurring

    Kaihao Zhang, Wenhan Luo, Yiran Zhong, Lin Ma, Bjorn Stenger, Wei Liu, and Hongdong Li. Deblurring by realistic blurring. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 2737–2746, 2020. 1

    work page 2020

  6. [4]

    Scale-recurrent network for deep image deblurring

    Xin Tao, Hongyun Gao, Xiaoyong Shen, Jue Wang, and Jiaya Jia. Scale-recurrent network for deep image deblurring. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 8174–8182, 2018. 1

    work page 2018

  7. [5]

    Deep multi-scale convolutional neural network for dynamic scene deblurring

    Seungjun Nah, Tae Hyun Kim, and Kyoung Mu Lee. Deep multi-scale convolutional neural network for dynamic scene deblurring. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 3883–3891, 2017. 1, 6

    work page 2017

  8. [6]

    Promp- tir: Prompting for all-in-one image restoration

    Vaishnav Potlapalli, Syed Waqas Zamir, Salman H Khan, and Fahad Shahbaz Khan. Promp- tir: Prompting for all-in-one image restoration. Advances in Neural Information Processing Systems, 36, 2024. 1, 7

    work page 2024

Show all 82 references
  1. [7]

    Airnet: Self-supervised affine registration for 3d medical images using neural networks

    Evelyn Chee and Zhenzhou Wu. Airnet: Self-supervised affine registration for 3d medical images using neural networks. arXiv preprint arXiv:1810.02583, 2018. 1, 7

    2018 arXiv

  2. [8]

    Deblurgan: Blind motion deblurring using conditional adversarial networks

    Orest Kupyn, V olodymyr Budzan, Mykola Mykhailych, Dmytro Mishkin, and Ji ˇrí Matas. Deblurgan: Blind motion deblurring using conditional adversarial networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 8183–8192, 2018. 1

    2018

  3. [9]

    Gan based image deblurring using dark channel prior

    Shuang Zhang, Ada Zhen, and Robert L Stevenson. Gan based image deblurring using dark channel prior. arXiv preprint arXiv:1903.00107, 2019. 1

    1903 arXiv

  4. [10]

    Diff-restorer: Unleashing visual prompts for diffusion-based universal image restoration

    Yuhong Zhang, Hengsheng Zhang, Xinning Chai, Zhengxue Cheng, Rong Xie, Li Song, and Wenjun Zhang. Diff-restorer: Unleashing visual prompts for diffusion-based universal image restoration. arXiv preprint arXiv:2407.03636, 2024. 1, 3, 6, 10

    2024 arXiv

  5. [11]

    Diffbir: Toward blind image restoration with generative diffusion prior

    Xinqi Lin, Jingwen He, Ziyan Chen, Zhaoyang Lyu, Bo Dai, Fanghua Yu, Yu Qiao, Wanli Ouyang, and Chao Dong. Diffbir: Toward blind image restoration with generative diffusion prior. In European Conference on Computer Vision, pages 430–448. Springer, 2025. 1, 3, 5, 6, 7

    2025

  6. [12]

    Diff-plugin: Revitalizing details for diffusion-based low-level tasks

    Yuhao Liu, Zhanghan Ke, Fang Liu, Nanxuan Zhao, and Rynson WH Lau. Diff-plugin: Revitalizing details for diffusion-based low-level tasks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4197–4208, 2024. 1, 3, 5, 7

    2024

  7. [13]

    Denoising diffusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. Advances in neural information processing systems, 33:6840–6851, 2020. 1, 3

    2020

  8. [14]

    High- resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. High- resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10684–10695, 2022. 1, 3

    2022

  9. [15]

    Dualgan: Unsupervised dual learning for image-to-image translation

    Zili Yi, Hao Zhang, Ping Tan, and Minglun Gong. Dualgan: Unsupervised dual learning for image-to-image translation. In Proceedings of the IEEE international conference on computer vision, pages 2849–2857, 2017. 2, 3

    2017

  10. [16]

    Fcl-gan: A lightweight and real-time baseline for unsupervised blind image deblurring

    Suiyi Zhao, Zhao Zhang, Richang Hong, Mingliang Xu, Yi Yang, and Meng Wang. Fcl-gan: A lightweight and real-time baseline for unsupervised blind image deblurring. In Proceedings of the 30th ACM International Conference on Multimedia, pages 6220–6229, 2022. 2, 3 22

    2022

  11. [17]

    Learning spatially-variant map models for non- blind image deblurring [c]

    DONG Jiangxin, S ROTH, and B SCHIELE. Learning spatially-variant map models for non- blind image deblurring [c]. the ieee. In CVF Conference on Computer Vision and Pattern Recognition, Nashville, USA, pages 4884–4893, 2021. 2, 3

    2021

  12. [18]

    Neural blind deconvolution using deep priors

    Dongwei Ren, Kai Zhang, Qilong Wang, Qinghua Hu, and Wangmeng Zuo. Neural blind deconvolution using deep priors. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 3341–3350, 2020. 2, 3

    2020

  13. [19]

    Uncertainty-aware variate decomposition for self-supervised blind image deblurring

    Runhua Jiang and Yahong Han. Uncertainty-aware variate decomposition for self-supervised blind image deblurring. In Proceedings of the 31st ACM International Conference on Multimedia, pages 252–260, 2023. 2, 3

    2023

  14. [20]

    Generative adversarial nets

    Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets. Advances in neural information processing systems, 27, 2014. 2, 3

    2014

  15. [21]

    Id- blau: Image deblurring by implicit diffusion-based reblurring augmentation

    Jia-Hao Wu, Fu-Jen Tsai, Yan-Tsung Peng, Chung-Chi Tsai, Chia-Wen Lin, and Yen-Yu Lin. Id- blau: Image deblurring by implicit diffusion-based reblurring augmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 25847–25856,

  16. [22]

    Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models

    JunnanLi, Dongxu Li, Silvio Savarese, and Steven Hoi. Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models. InInternational conference on machine learning, pages 19730–19742. PMLR, 2023. 2, 4

    2023

  17. [23]

    Deep unsuper- vised learning using nonequilibrium thermodynamics

    Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep unsuper- vised learning using nonequilibrium thermodynamics. In International conference on machine learning, pages 2256–2265. PMLR, 2015. 3

    2015

  18. [24]

    Generative modeling by estimating gradients of the data distribution

    Yang Song and Stefano Ermon. Generative modeling by estimating gradients of the data distribution. Advances in neural information processing systems, 32, 2019. 3

    2019

  19. [25]

    Denoising diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. arXiv: Learning,arXiv: Learning, Oct 2020. 3, 7

    2020

  20. [26]

    Ip-adapter: Text compatible image prompt adapter for text-to-image diffusion models

    Hu Ye, Jun Zhang, Sibo Liu, Xiao Han, and Wei Yang. Ip-adapter: Text compatible image prompt adapter for text-to-image diffusion models. arXiv preprint arXiv:2308.06721, 2023. 3, 5

    2023 arXiv

  21. [27]

    Autostudio: Crafting consistent subjects in multi-turn interactive image generation

    Junhao Cheng, Xi Lu, Hanhui Li, Khun Loun Zai, Baiqiao Yin, Yuhao Cheng, Yiqiang Yan, and Xiaodan Liang. Autostudio: Crafting consistent subjects in multi-turn interactive image generation. arXiv preprint arXiv:2406.01388, 2024. 3

    2024 arXiv

  22. [28]

    Object isolated attention for consistent story visualization

    Xiangyang Luo, Junhao Cheng, Yifan Xie, Xin Zhang, Tao Feng, Zhou Liu, Fei Ma, and Fei Yu. Object isolated attention for consistent story visualization. arXiv preprint arXiv:2503.23353,

    arXiv

  23. [29]

    Laion- 5b: An open large-scale dataset for training next generation image-text models

    Christoph Schuhmann, Romain Beaumont, Richard Vencu, Cade Gordon, Ross Wightman, Mehdi Cherti, Theo Coombes, Aarush Katta, Clayton Mullis, Mitchell Wortsman, et al. Laion- 5b: An open large-scale dataset for training next generation image-text models. Advances in Neural Inform...

    2022

  24. [30]

    Open images v5 text annotation and yet another mask text spotter

    Ilya Krylov, Sergei Nosov, and Vladislav Sovrasov. Open images v5 text annotation and yet another mask text spotter. In Asian Conference on Machine Learning, pages 379–389. PMLR,

  25. [31]

    Diffusion models beat gans on image synthesis

    Prafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis. Advances in neural information processing systems, 34:8780–8794, 2021. 3

    2021

  26. [32]

    Classifier-free diffusion guidance

    Jonathan Ho and Tim Salimans. Classifier-free diffusion guidance. arXiv preprint arXiv:2207.12598, 2022. 3, 7

    2022 arXiv

  27. [33]

    Exploiting deep generative prior for versatile image restoration and manipulation

    Xingang Pan, Xiaohang Zhan, Bo Dai, Dahua Lin, Chen Change Loy, and Ping Luo. Exploiting deep generative prior for versatile image restoration and manipulation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(11):7474–7489, 2021. 3 23

    2021

  28. [34]

    Improving diffusion models for inverse problems using manifold constraints

    Hyungjin Chung, Byeongsu Sim, Dohoon Ryu, and Jong Chul Ye. Improving diffusion models for inverse problems using manifold constraints. Advances in Neural Information Processing Systems, 35:25683–25696, 2022. 3

    2022

  29. [35]

    Zero-shot image restoration using denoising diffusion null-space model

    Yinhuai Wang, Jiwen Yu, and Jian Zhang. Zero-shot image restoration using denoising diffusion null-space model. arXiv preprint arXiv:2212.00490, 2022. 3

    2022 arXiv

  30. [36]

    Diffu- sion posterior sampling for general noisy inverse problems

    Hyungjin Chung, Jeongsol Kim, Michael T Mccann, Marc L Klasky, and Jong Chul Ye. Diffu- sion posterior sampling for general noisy inverse problems. arXiv preprint arXiv:2209.14687,

    arXiv

  31. [37]

    Denoising diffusion restoration models

    Bahjat Kawar, Michael Elad, Stefano Ermon, and Jiaming Song. Denoising diffusion restoration models. Advances in Neural Information Processing Systems, 35:23593–23606, 2022. 3

    2022

  32. [38]

    Srdiff: Single image super-resolution with diffusion probabilistic models

    Haoying Li, Yifan Yang, Meng Chang, Shiqi Chen, Huajun Feng, Zhihai Xu, Qi Li, and Yueting Chen. Srdiff: Single image super-resolution with diffusion probabilistic models. Neurocomputing, 479:47–59, 2022. 3

    2022

  33. [39]

    Restoring vision in adverse weather conditions with patch-based denoising diffusion models

    Ozan Özdenizci and Robert Legenstein. Restoring vision in adverse weather conditions with patch-based denoising diffusion models. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(8):10346–10357, 2023. 3

    2023

  34. [40]

    Mul- tiscale structure guided diffusion for image deblurring

    Mengwei Ren, Mauricio Delbracio, Hossein Talebi, Guido Gerig, and Peyman Milanfar. Mul- tiscale structure guided diffusion for image deblurring. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 10721–10733, 2023. 3

    2023

  35. [41]

    GibbsDDRM: A partially collapsed gibbs sampler for solving blind inverse problems with denoising diffusion restoration

    Naoki Murata, Koichi Saito, Chieh-Hsin Lai, Yuhta Takida, Toshimitsu Uesaka, Yuki Mitsufuji, and Stefano Ermon. GibbsDDRM: A partially collapsed gibbs sampler for solving blind inverse problems with denoising diffusion restoration. In International Conference on Machine Learni...

    2023

  36. [42]

    Pre-trained image processing transformer

    Hanting Chen, Yunhe Wang, Tianyu Guo, Chang Xu, Yiping Deng, Zhenhua Liu, Siwei Ma, Chunjing Xu, Chao Xu, and Wen Gao. Pre-trained image processing transformer. arXiv, 2021. 3

    2021

  37. [43]

    Parallel diffusion models of operator and image for blind inverse problems

    Hyungjin Chung, Jeongsol Kim, Sehui Kim, and Jong Chul Ye. Parallel diffusion models of operator and image for blind inverse problems. IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023. 3

    2023

  38. [44]

    Exploiting diffusion prior for real-world image super-resolution

    Jianyi Wang, Zongsheng Yue, Shangchen Zhou, Kelvin CK Chan, and Chen Change Loy. Exploiting diffusion prior for real-world image super-resolution. International Journal of Computer Vision, pages 1–21, 2024. 3

    2024

  39. [45]

    Controlling vision-language models for universal image restoration

    Ziwei Luo, Fredrik K Gustafsson, Zheng Zhao, Jens Sjölund, and Thomas B Schön. Controlling vision-language models for universal image restoration. arXiv preprint arXiv:2310.01018, 3(8),

    arXiv

  40. [46]

    Deblurdiff: Real-world image deblurring with generative diffusion models, 2025

    Lingshun Kong, Jiawei Zhang, Dongqing Zou, Jimmy Ren, Xiaohe Wu, Jiangxin Dong, and Jinshan Pan. Deblurdiff: Real-world image deblurring with generative diffusion models, 2025. 3

    2025

  41. [48]

    Realistic blur synthesis for learning image deblurring

    Jaesung Rim, Geonung Kim, Jungeon Kim, Junyong Lee, Seungyong Lee, and Sunghyun Cho. Realistic blur synthesis for learning image deblurring. In European conference on computer vision, pages 487–503. Springer, 2022. 3

    2022

  42. [49]

    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. 3

    2022

  43. [50]

    Rethink- ing coarse-to-fine approach in single image deblurring

    Sung-Jin Cho, Seo-Won Ji, Jun-Pyo Hong, Seung-Won Jung, and Sung-Jea Ko. Rethink- ing coarse-to-fine approach in single image deblurring. In Proceedings of the IEEE/CVF international conference on computer vision, pages 4641–4650, 2021. 3 24

    2021

  44. [51]

    Learning degradation representations for image deblurring

    Dasong Li, Yi Zhang, Ka Chun Cheung, Xiaogang Wang, Hongwei Qin, and Hongsheng Li. Learning degradation representations for image deblurring. In European conference on computer vision, pages 736–753. Springer, 2022. 3

    2022

  45. [52]

    Intriguing findings of frequency selection for image deblurring

    Xintian Mao, Yiming Liu, Fengze Liu, Qingli Li, Wei Shen, and Yan Wang. Intriguing findings of frequency selection for image deblurring. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pages 1905–1913, 2023. 3

    1905

  46. [53]

    Stripformer: Strip transformer for fast image deblurring

    Fu-Jen Tsai, Yan-Tsung Peng, Yen-Yu Lin, Chung-Chi Tsai, and Chia-Wen Lin. Stripformer: Strip transformer for fast image deblurring. In European conference on computer vision, pages 146–162. Springer, 2022. 3

    2022

  47. [54]

    Efficient frequency domain-based transformers for high-quality image deblurring

    Lingshun Kong, Jiangxin Dong, Jianjun Ge, Mingqiang Li, and Jinshan Pan. Efficient frequency domain-based transformers for high-quality image deblurring. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5886–5895, 2023. 3

    2023

  48. [55]

    Image deblurring by exploring in- depth properties of transformer

    Pengwei Liang, Junjun Jiang, Xianming Liu, and Jiayi Ma. Image deblurring by exploring in- depth properties of transformer. IEEE Transactions on Neural Networks and Learning Systems,

  49. [56]

    Deblurdinat: A lightweight and effective transformer for image deblurring

    Hanzhou Liu, Binghan Li, Chengkai Liu, and Mi Lu. Deblurdinat: A lightweight and effective transformer for image deblurring. arXiv preprint arXiv:2403.13163, 2024. 3

    2024 arXiv

  50. [57]

    Self-supervised non-uniform kernel estimation with flow-based motion prior for blind im- age deblurring

    Zhenxuan Fang, Fangfang Wu, Weisheng Dong, Xin Li, Jinjian Wu, and Guangming Shi. Self-supervised non-uniform kernel estimation with flow-based motion prior for blind im- age deblurring. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, page...

    2023

  51. [58]

    Motion-adaptive separable collaborative filters for blind motion deblurring

    Chengxu Liu, Xuan Wang, Xiangyu Xu, Ruhao Tian, Shuai Li, Xueming Qian, and Ming- Hsuan Yang. Motion-adaptive separable collaborative filters for blind motion deblurring. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 25595–25605, 2024. 3

    2024

  52. [59]

    Hierarchical integration diffusion model for realistic image deblurring

    Zheng Chen, Yulun Zhang, Ding Liu, Jinjin Gu, Linghe Kong, Xin Yuan, et al. Hierarchical integration diffusion model for realistic image deblurring. Advances in neural information processing systems, 36, 2024. 3

    2024

  53. [60]

    Fast diffusion em: a diffusion model for blind inverse problems with application to deconvolution

    Charles Laroche, Andrés Almansa, and Eva Coupete. Fast diffusion em: a diffusion model for blind inverse problems with application to deconvolution. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pages 5271–5281, 2024. 3

    2024

  54. [61]

    Learning spatially-variant map models for non-blind image deblurring

    Jiangxin Dong, Stefan Roth, and Bernt Schiele. Learning spatially-variant map models for non-blind image deblurring. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 4886–4895, 2021. 3

    2021

  55. [62]

    Auto-encoding variational bayes

    Diederik P Kingma. Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114, 2013. 3

    2013 arXiv

  56. [63]

    Blip-diffusion: Pre-trained subject representation for controllable text-to-image generation and editing

    DongxuLi, Junnan Li, and Steven Hoi. Blip-diffusion: Pre-trained subject representation for controllable text-to-image generation and editing. Advances in Neural Information Processing Systems, 36, 2024. 4, 7

    2024

  57. [64]

    Theatergen: Character management with llm for consistent multi-turn image generation

    Junhao Cheng, Baiqiao Yin, Kaixin Cai, Minbin Huang, Hanhui Li, Yuxin He, Xi Lu, Yue Li, Yifei Li, Yuhao Cheng, et al. Theatergen: Character management with llm for consistent multi-turn image generation. arXiv preprint arXiv:2404.18919, 2024. 5

    2024 arXiv

  58. [65]

    Con- trolnext: Powerful and efficient control for image and video generation

    Bohao Peng, Jian Wang, Yuechen Zhang, Wenbo Li, Ming-Chang Yang, and Jiaya Jia. Con- trolnext: Powerful and efficient control for image and video generation. arXiv preprint arXiv:2408.06070, 2024. 5

    2024 arXiv

  59. [66]

    Adding conditional control to text-to-image diffusion models

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 3836–3847, 2023. 5 25

    2023

  60. [67]

    T2i-adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models

    Chong Mou, Xintao Wang, Liangbin Xie, Yanze Wu, Jian Zhang, Zhongang Qi, and Ying Shan. T2i-adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models. In Proceedings of the AAAI Conference on Artificial Intelligence, 2024. 5

    2024

  61. [68]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. In International conference on machine learning, pa...

    2021

  62. [69]

    Artadapter: Text-to-image style transfer using multi-level style encoder and explicit adaptation

    Dar-Yen Chen, Hamish Tennent, and Ching-Wen Hsu. Artadapter: Text-to-image style transfer using multi-level style encoder and explicit adaptation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8619–8628, 2024. 5

    2024

  63. [70]

    Unirestore: Unified perceptual and task-oriented image restoration model using diffusion prior

    I Chen, Wei-Ting Chen, Yu-Wei Liu, Yuan-Chun Chiang, Sy-Yen Kuo, Ming-Hsuan Yang, et al. Unirestore: Unified perceptual and task-oriented image restoration model using diffusion prior. arXiv preprint arXiv:2501.13134, 2025. 6, 10

    2025 arXiv

  64. [72]

    Ntire 2019 challenge on video deblurring and super-resolution: Dataset and study

    Seungjun Nah, Sungyong Baik, Seokil Hong, Gyeongsik Moon, Sanghyun Son, Radu Tim- ofte, and Kyoung Mu Lee. Ntire 2019 challenge on video deblurring and super-resolution: Dataset and study. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition wor...

    2019

  65. [73]

    Maniqa: Multi-dimension attention network for no-reference image quality assessment

    Sidi Yang, Tianhe Wu, Shuwei Shi, Shanshan Lao, Yuan Gong, Mingdeng Cao, Jiahao Wang, and Yujiu Yang. Maniqa: Multi-dimension attention network for no-reference image quality assessment. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, page...

    2022

  66. [74]

    Blind image quality as- sessment via vision-language correspondence: A multitask learning perspective

    Weixia Zhang, Guangtao Zhai, Ying Wei, Xiaokang Yang, and Kede Ma. Blind image quality as- sessment via vision-language correspondence: A multitask learning perspective. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 14071–14081,

  67. [75]

    Musiq: Multi-scale image quality transformer

    Junjie Ke, Qifei Wang, Yilin Wang, Peyman Milanfar, and Feng Yang. Musiq: Multi-scale image quality transformer. In Proceedings of the IEEE/CVF international conference on computer vision, pages 5148–5157, 2021. 6

    2021

  68. [76]

    Exploring clip for assessing the look and feel of images

    Jianyi Wang, Kelvin CK Chan, and Chen Change Loy. Exploring clip for assessing the look and feel of images. In Proceedings of the AAAI Conference on Artificial Intelligence, 2023. 6

    2023

  69. [77]

    Decoupled weight decay regularization

    I Loshchilov. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101, 2017. 7

    2017 arXiv

  70. [78]

    Adarevd: Adaptive patch exiting reversible decoder pushes the limit of image deblurring

    Qingli Li Xintian Mao and Yan Wang. Adarevd: Adaptive patch exiting reversible decoder pushes the limit of image deblurring. In Proc. CVPR, 2024. 7

    2024

  71. [79]

    Efficient frequency domain-based transformers for high-quality image deblurring, 2022

    Lingshun Kong, Jiangxin Dong, Mingqiang Li, Jianjun Ge, and Jinshan Pan. Efficient frequency domain-based transformers for high-quality image deblurring, 2022. 7

    2022

  72. [80]

    Hierar- chical integration diffusion model for realistic image deblurring

    Zheng Chen, Yulun Zhang, Liu Ding, Xia Bin, Jinjin Gu, Linghe Kong, and Xin Yuan. Hierar- chical integration diffusion model for realistic image deblurring. In NeurIPS, 2023. 7

    2023

  73. [81]

    Designing a better asymmetric vqgan for stablediffusion

    Zixin Zhu, Xuelu Feng, Dongdong Chen, Jianmin Bao, Le Wang, Yinpeng Chen, Lu Yuan, and Gang Hua. Designing a better asymmetric vqgan for stablediffusion. arXiv preprint arXiv:2306.04632, 2023. 10

    2023 arXiv

  74. [82]

    The unrea- sonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shechtman, and Oliver Wang. The unrea- sonable effectiveness of deep features as a perceptual metric. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 586–595, 2018. 10 26

    2018

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.