pith. machine review for the scientific record. sign in

arxiv: 2603.16446 · v3 · submitted 2026-03-17 · 💻 cs.CV

Unified Removal of Raindrops and Reflections: A New Benchmark and A Novel Pipeline

Xingyu Liu , Zewei He , Yu Chen , Chunyu Zhu , Zixuan Chen , Xing Luo , Zhe-Ming Lu This is my paper

Pith reviewed 2026-05-15 09:51 UTC · model grok-4.3

classification 💻 cs.CV
keywords raindrop removalreflection removalimage restorationdiffusion modelsadverse weatherbenchmark datasetcomputer vision
0
0 comments X

The pith

A new diffusion framework and real-shot dataset enable simultaneous removal of raindrops and reflections from images taken through glass.

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

The paper formally defines the unified removal of raindrops and reflections task and builds the RDRF dataset of real paired images to serve as a benchmark. It then introduces DiffUR³, a diffusion-based pipeline that applies generative priors to restore visibility when both degradations appear together. Prior single-task or all-in-one methods left this composite case unaddressed, so the work targets a frequent practical failure mode in photography and outdoor vision systems. If the approach holds, restored images would support downstream tasks such as object detection on rainy windshields without separate preprocessing steps.

Core claim

The central claim is that the RDRF benchmark combined with the DiffUR³ diffusion framework, equipped with several target designs, successfully removes both raindrops and reflections at once and reaches state-of-the-art performance on the new benchmark as well as on challenging in-the-wild images, where earlier de-raindrop, de-reflection, and all-in-one models fall short.

What carries the argument

DiffUR³, a diffusion-based restoration pipeline that leverages generative priors to jointly address raindrop and reflection degradations through targeted architectural designs.

If this is right

  • Clearer output images from cameras behind windshields or windows on rainy days without manual intervention.
  • A public paired dataset that future methods can use to measure progress on combined degradations.
  • Improved input quality for downstream computer-vision pipelines that currently fail when both raindrops and reflections are present.
  • Outperformance over separate de-raindrop and de-reflection models when the two problems co-occur.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same diffusion-prior strategy could be tested on other paired weather degradations such as rain streaks plus fog.
  • Autonomous-driving perception stacks might adopt the pipeline directly for windshield-mounted cameras once the dataset is released.
  • The benchmark could reveal whether purely generative approaches generalize better than supervised regression methods to rare but critical edge cases.

Load-bearing premise

The RDRF dataset and its diffusion prior capture enough of the real-world variety of raindrop-plus-reflection combinations that the trained model will see after deployment.

What would settle it

A new test set of images containing raindrops and reflections on glass surfaces, lighting, or angles absent from RDRF would show whether restoration quality drops sharply compared with results on the original benchmark.

Figures

Figures reproduced from arXiv: 2603.16446 by Chunyu Zhu, Xing Luo, Xingyu Liu, Yu Chen, Zewei He, Zhe-Ming Lu, Zixuan Chen.

Figure 1
Figure 1. Figure 1: We compare our DiffUR3 pipeline with other methods on low-quality images with raindrops and reflections from our newly collected real-world benchmark. Specif￾ically, (c) DAI [14] is designed for reflection removal, (d) A cascaded method, and (e) A re-trained all-in-one method (i.e., Histoformer [36] and † indicates re-trained on our dataset), (f) Our DiffUR3 pipeline jointly removes both degradations in a … view at source ↗
Figure 2
Figure 2. Figure 2: (a) Sketch diagram and actual equipment of our image acquisition platform. To suppress shutter-induced micro-vibrations which may potentially induce image mis￾alignment, we implement a wireless triggering mechanism. It comprises a remote con￾troller and a camera-mounted signal receiver, enabling contact-free shutter operation. (b) The data collection pipeline for our RDRF dataset. é denotes light occlusion… view at source ↗
Figure 3
Figure 3. Figure 3: Our RDRF dataset comprises a diverse collection of scenes, each contains a ground truth and multiple low-quality images. As illustrated in this figure, the clean ground truths are highlighted in red boxes, while corresponding low-quality images are arranged around. We divide it into the training and testing subsets, ensuring no overlapping samples between them. Please zoom in on screen for a better view. o… view at source ↗
Figure 4
Figure 4. Figure 4: (a) Overall pipeline of our DiffUR3 framework. Given a low-quality image Ilq, the restoration stage removes the undesired degradation to obtain the initial result Is. Both Ilq and Is are fed into the next stage as the condition images. We inject the effective condition information through a control branch, which outputs control signals for the noise prediction U-Net. (b) Details of the Modulate&Gate module… view at source ↗
Figure 5
Figure 5. Figure 5: (b)). The stable diffusion model lacks the ability to correct these kinds of distortions. To deal with this issue and improve the fidelity of the generated results, we train an additional fidelity encoder (FE) inspired by [1]. The FE is proposed to extract multi-scale features from the initial result Is and the LQ image Ilq, which are not affected by the down-sampling, for preserving local structural seman… view at source ↗
Figure 7
Figure 7. Figure 7: In-depth analysis on the function of our Modulate&Gate (M&G) module. (a) Baseline (b) w/ CFW (c) w/ FE [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 10
Figure 10. Figure 10: Color correction results with different methods. 5.3 Comparisons with state-of-the-art methods Since this work is the first exploration for unified removal of raindrops and reflections (UR3 ) task. There are no prior methods. We employ two classical raindrop removal methods (i.e., AGAN [29], UMAN [33]), three classical reflec￾tion removal methods (i.e., RDNet [45], DSIT [17], DAI [14]), four cascaded meth… view at source ↗
Figure 11
Figure 11. Figure 11: Visual results of various methods on our RDRF-testing. Superscript † means this method is re-trained on our RDRF-traing dataset. Please check and zoom in on screen for a better view. (a) LQ (b) AGAN+RDNet (c) RDNet+AGAN (d) Histoformer † (e) DiffUIR† (f) Stage I (g) Ours [PITH_FULL_IMAGE:figures/full_fig_p014_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Visual results of various methods on our RDRF-wild dataset. Please check and zoom in on screen for a better view. in rainy weather are very challenging, we capture some testing images to form a RDRF-wild dataset [PITH_FULL_IMAGE:figures/full_fig_p014_12.png] view at source ↗
read the original abstract

When capturing images through glass surfaces or windshields on rainy days, raindrops and reflections frequently co-occur to significantly reduce the visibility of captured images. This practical problem lacks attention and needs to be resolved urgently. Prior de-raindrop, de-reflection, and all-in-one models have failed to address this composite degradation. To this end, we first formally define the unified removal of raindrops and reflections (UR$^3$) task for the first time and construct a real-shot dataset, namely RainDrop and ReFlection (RDRF), which provides a new benchmark with substantial, high-quality, diverse image pairs. Then, we propose a novel diffusion-based framework (i.e., DiffUR$^3$) with several target designs to address this challenging task. By leveraging the powerful generative prior, DiffUR$^3$ successfully removes both types of degradations. Extensive experiments demonstrate that our method achieves state-of-the-art performance on our benchmark and on challenging in-the-wild images. The RDRF dataset and the codes will be made public upon acceptance.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper formally defines the unified removal of raindrops and reflections (UR³) task, constructs the real-shot RDRF dataset as a new benchmark with paired images, and proposes DiffUR³, a diffusion-based framework that uses generative priors to jointly remove both degradations, claiming state-of-the-art quantitative and qualitative results on the RDRF benchmark and challenging in-the-wild images.

Significance. If the empirical margins hold after verification, the work supplies the first dedicated benchmark and method for a practically common composite degradation, with the diffusion prior providing a plausible route to handling the joint distribution of raindrops and reflections. The release of the dataset and code would further strengthen its utility for the community.

major comments (2)
  1. [Dataset Construction] Dataset Construction section: the RDRF dataset is presented without quantitative coverage metrics (e.g., histograms or statistics on raindrop density, reflection contrast, glass curvature, or scene diversity), which directly bears on the representativeness assumption required to support both the benchmark SOTA numbers and the in-the-wild generalization claims.
  2. [Experiments] Experiments section: full training hyperparameters, exact baseline re-implementations, and statistical significance tests (e.g., paired t-tests or confidence intervals on the reported PSNR/SSIM margins) are omitted, preventing independent confirmation that the observed improvements are robust rather than artifacts of evaluation protocol.
minor comments (2)
  1. [Method] Clarify the precise architectural modifications to the diffusion backbone (e.g., conditioning mechanisms for the composite degradation) in the method description.
  2. [Conclusion] Add a limitations paragraph discussing failure cases under extreme lighting or dense raindrop overlap.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment below and will incorporate revisions to strengthen the paper.

read point-by-point responses

  1. Referee: [Dataset Construction] Dataset Construction section: the RDRF dataset is presented without quantitative coverage metrics (e.g., histograms or statistics on raindrop density, reflection contrast, glass curvature, or scene diversity), which directly bears on the representativeness assumption required to support both the benchmark SOTA numbers and the in-the-wild generalization claims.

    Authors: We agree that quantitative coverage metrics would better substantiate the representativeness of the RDRF dataset. In the revised manuscript, we will expand the Dataset Construction section to include histograms and summary statistics on raindrop density, reflection contrast levels, glass curvature variations, and scene diversity (e.g., indoor/outdoor, lighting conditions). These metrics were recorded during curation and will be presented to support the benchmark validity and generalization claims. revision: yes

  2. Referee: [Experiments] Experiments section: full training hyperparameters, exact baseline re-implementations, and statistical significance tests (e.g., paired t-tests or confidence intervals on the reported PSNR/SSIM margins) are omitted, preventing independent confirmation that the observed improvements are robust rather than artifacts of evaluation protocol.

    Authors: We acknowledge that these implementation details were omitted. In the revised version, we will add a dedicated subsection in Experiments detailing all training hyperparameters (learning rate, batch size, diffusion steps, etc.), exact re-implementation procedures for baselines (including any adaptations to our task), and statistical analyses such as paired t-tests with p-values and 95% confidence intervals on the PSNR/SSIM margins to confirm robustness. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical SOTA claims rest on independent dataset construction and experimental measurement

full rationale

The paper introduces the UR^3 task definition, constructs the RDRF real-shot paired dataset as an external benchmark, proposes the DiffUR^3 diffusion framework, and reports performance metrics obtained by running the trained model on held-out test images from that dataset. No equations, fitted parameters, or self-citations are invoked in a load-bearing manner that would make the reported metrics equivalent to quantities defined by the same inputs. The evaluation is therefore self-contained and falsifiable against the released dataset.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The claim rests on the representativeness of the newly collected RDRF pairs and on the standard generative prior of diffusion models being sufficient to invert the joint raindrop-plus-reflection degradation.

free parameters (1)
  • Diffusion network weights
    All parameters of the U-Net backbone and conditioning modules are fitted to the RDRF training split.
axioms (1)
  • domain assumption Diffusion models trained on paired restoration data can invert composite real-world degradations
    Invoked when the authors state that leveraging the generative prior successfully removes both raindrops and reflections.

pith-pipeline@v0.9.0 · 5503 in / 1339 out tokens · 56615 ms · 2026-05-15T09:51:21.348729+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

51 extracted references · 51 canonical work pages · 2 internal anchors

  1. [1]

    In: NeurIPS

    Chang, Z., Weng, S., Zhang, P., Li, Y., Li, S., Shi, B.: L-CAD: Language-based Col- orization with Any-level Descriptions using Diffusion Priors. In: NeurIPS. vol. 36, pp. 1–13 (2023) 9

    work page 2023

  2. [2]

    In: CVPR

    Chen, I., Chen, W.T., Liu, Y.W., Chiang, Y.C., Kuo, S.Y., Yang, M.H., et al.: Unirestore: Unified perceptual and task-oriented image restoration model using diffusion prior. In: CVPR. pp. 17969–17979 (2025) 12, 13

    work page 2025

  3. [3]

    In: CVPR (2025) 8

    Chen, J., Pan, J., Dong, J.: FaithDiff: Unleashing Diffusion Priors for Faithful Image Super-resolution. In: CVPR (2025) 8

    work page 2025

  4. [4]

    In: ICCV

    Chen, S., Ye, T., Bai, J., Chen, E., Shi, J., Zhu, L.: Sparse Sampling Trans- former with Uncertainty-Driven Ranking for Unified Removal of Raindrops and Rain Streaks. In: ICCV. pp. 13060–13071 (2023) 3

    work page 2023

  5. [5]

    In: ECCV

    Chen, S., Ye, T., Zhang, K., Xing, Z., Lin, Y., Zhu, L.: Teaching Tailored to Talent: Adverse Weather Restoration via Prompt Pool and Depth-Anything Constraint. In: ECCV. pp. 95–115 (2024) 3

    work page 2024

  6. [6]

    In: CVPR

    Chen, X., Li, H., Li, M., Pan, J.: Learning A Sparse Transformer Network for Effective Image Deraining. In: CVPR. pp. 5896–5905 (2023) 6, 7, 10

    work page 2023

  7. [7]

    In: CVPR (2026) 3

    Chen, Y., He, Z., Liu, X., Chen, Z., Lu, Z.: Gfrrn: Explore the gaps in single image reflection removal. In: CVPR (2026) 3

    work page 2026

  8. [8]

    In: CVPR

    Choi, J., Lee, J., Shin, C., Kim, S., Kim, H., Yoon, S.: Perception Prioritized Training of Diffusion Models. In: CVPR. pp. 11462–11471 (2022) 10

    work page 2022

  9. [9]

    In: ICCV

    Eigen, D., Krishnan, D., Fergus, R.: Restoring an image taken through a window covered with dirt or rain. In: ICCV. pp. 633–640 (2013) 3

    work page 2013

  10. [10]

    In: ICCV

    Fan, Q., Yang, J., Hua, G., Chen, B., Wipf, D.: A Generic Deep Architecture for Single Image Reflection Removal and Image Smoothing. In: ICCV. pp. 3238–3247 (2017) 3

    work page 2017

  11. [11]

    Communi- cations of the ACM24(6), 381–395 (1981) 6

    Fischler, M.A., Bolles, R.C.: Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography. Communi- cations of the ACM24(6), 381–395 (1981) 6

    work page 1981

  12. [12]

    In: CVPR

    He, T., Zhang, Z., Zhang, H., Zhang, Z., Xie, J., Li, M.: Bag of tricks for image classification with convolutional neural networks. In: CVPR. pp. 558–567 (2019) 10

    work page 2019

  13. [13]

    In: ECCV

    Hong,Y.,Zhong,H.,Weng,S.,Liang,J.,Shi,B.:L-DiffER:SingleImageReflection Removal with Language-Based Diffusion Model. In: ECCV. pp. 58–76 (2024) 3

    work page 2024

  14. [14]

    In: AAAI (2026) 2, 3, 4, 12, 13

    Hu, J., Yang, C., Zhou, Z., Fang, J., Yang, X., Tian, Q., Shen, W.: Dereflection Any Image with Diffusion Priors and Diversified Data. In: AAAI (2026) 2, 3, 4, 12, 13

    work page 2026

  15. [15]

    In: NeurIPS

    Hu, Q., Guo, X.: Trash or Treasure? An Interactive Dual-Stream Strategy for Single Image Reflection Separation. In: NeurIPS. vol. 30, pp. 24683–24694 (2021) 3

    work page 2021

  16. [16]

    In: ICCV

    Hu, Q., Guo, X.: Single Image Reflection Separation via Component Synergy. In: ICCV. pp. 13092–13101 (2023) 3

    work page 2023

  17. [17]

    In: NeurIPS

    Hu, Q., Wang, H., Guo, X.: Single image reflection separation via dual-stream interactive transformers. In: NeurIPS. vol. 37, pp. 55228–55248 (2024) 2, 3, 12, 13

    work page 2024

  18. [18]

    In: ECCV (2024) 4, 6

    Jin, Y., Li, X., Wang, J., Zhang, Y., Zhang, M.: Raindrop Clarity: A Dual-Focused Dataset for Day and Night Raindrop Removal. In: ECCV (2024) 4, 6

    work page 2024

  19. [19]

    In: ICCV

    Ke, J., Wang, Q., Wang, Y., Milanfar, P., Yang, F.: Musiq: Multi-scale image quality transformer. In: ICCV. pp. 5148–5157 (2021) 10 16 X. Liu et al

    work page 2021

  20. [20]

    Auto-Encoding Variational Bayes

    Kingma, D.P., Welling, M.: Auto-encoding variational bayes. arXiv1312.6114 (2013) 7, 9

    work page internal anchor Pith review Pith/arXiv arXiv 2013

  21. [21]

    In: CVPR

    Li, C., Yang, Y., He, K., Lin, S., Hopcroft, J.E.: Single image reflection removal through cascaded refinement. In: CVPR. pp. 3562–3571 (2020) 3

    work page 2020

  22. [22]

    Li,Y.,Monno,Y.,Okutomi,M.:Dual-PixelRaindropRemoval.IEEETransactions on Pattern Analysis and Machine Intelligence46(12), 10748–10762 (2024) 4

    work page 2024

  23. [23]

    In: ECCV

    Lin, X., He, J., Chen, Z., Lyu, Z., Dai, B., Yu, F., Qiao, Y., Ouyang, W., Dong, C.: DiffBIR: Toward Blind Image Restoration with Generative Diffusion Prior. In: ECCV. pp. 430–448 (2024) 7, 8, 11

    work page 2024

  24. [24]

    Decoupled Weight Decay Regularization

    Loshchilov, I., Hutter, F., et al.: Fixing weight decay regularization in adam. ArXiv 1711.05101(2017) 10

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  25. [25]

    Interna- tional journal of computer vision60(2), 91–110 (2004) 6

    Lowe, D.G.: Distinctive image features from scale-invariant keypoints. Interna- tional journal of computer vision60(2), 91–110 (2004) 6

    work page 2004

  26. [26]

    In: ICML

    Nichol, A.Q., Dhariwal, P.: Improved denoising diffusion probabilistic models. In: ICML. pp. 8162–8171 (2021) 10

    work page 2021

  27. [27]

    IEEE Transactions on Pattern Analysis and Machine Intelligence45(8), 10346–10357 (2023) 3, 8, 10, 12, 13

    Özdenizci, O., Legenstein, R.: Restoring Vision in Adverse Weather Conditions With Patch-Based Denoising Diffusion Models. IEEE Transactions on Pattern Analysis and Machine Intelligence45(8), 10346–10357 (2023) 3, 8, 10, 12, 13

    work page 2023

  28. [28]

    IEEE International Conference on Robotics and Automation pp

    Porav, H., Bruls, T., Newman, P.: I can see clearly now: Image restoration via de-raining. IEEE International Conference on Robotics and Automation pp. 7087– 7093 (2019) 4

    work page 2019

  29. [29]

    In: CVPR

    Qian, R., Tan, R.T., Yang, W., Su, J., Liu, J.: Attentive Generative Adversarial Network for Raindrop Removal from a Single Image. In: CVPR. pp. 2482–2491 (2018) 2, 3, 4, 6, 12, 13

    work page 2018

  30. [30]

    In: CVPR

    Quan, R., Yu, X., Liang, Y., Yang, Y.: Removing Raindrops and Rain Streaks in One Go. In: CVPR. pp. 9143–9152 (2021) 3, 4, 6

    work page 2021

  31. [31]

    In: ICCV

    Quan, Y., Deng, S., Chen, Y., Ji, H.: Deep learning for seeing through window with raindrops. In: ICCV. pp. 2463–2471 (2019) 1, 2

    work page 2019

  32. [32]

    In: CVPR

    Rombach, R., Blattmann, A., Lorenz, D., Esser, P., Ommer, B.: High-resolution image synthesis with latent diffusion models. In: CVPR. pp. 10684–10695 (2022) 7, 10

    work page 2022

  33. [33]

    IEEE Transactions on Image Processing30, 4828–4839 (2021) 2, 12, 13

    Shao, M.W., Li, L., Meng, D.Y., Zuo, W.M.: Uncertainty Guided Multi-Scale At- tention Network for Raindrop Removal from a Single Image. IEEE Transactions on Image Processing30, 4828–4839 (2021) 2, 12, 13

    work page 2021

  34. [34]

    In: IEEE Symposium Series on Computa- tional Intelligence

    Soboleva, V., Shipitko, O.: Raindrops on Windshield: Dataset and Lightweight Gradient-Based Detection Algorithm. In: IEEE Symposium Series on Computa- tional Intelligence. pp. 1–7 (2021) 4

    work page 2021

  35. [35]

    In: CVPR

    Su, S., Yan, Q., Zhu, Y., Zhang, C., Ge, X., Sun, J., Zhang, Y.: Blindly Assess Image Quality in the Wild Guided by a Self-Adaptive Hyper Network. In: CVPR. pp. 3664–3673 (2020) 10

    work page 2020

  36. [36]

    In: ECCV (2024) 2, 3, 6, 12, 13

    Sun,S.,Ren,W.,Gao,X.,Wang,R.,Cao,X.:RestoringImagesinAdverseWeather Conditions via Histogram Transformer. In: ECCV (2024) 2, 3, 6, 12, 13

    work page 2024

  37. [37]

    In: NeurIPS

    Vaswani,A.,Shazeer,N.,Parmar,N.,Uszkoreit,J.,Jones,L.,Gomez,A.N.,Kaiser, L.u., Polosukhin, I.: Attention is all you need. In: NeurIPS. vol. 30 (2017) 8

    work page 2017

  38. [38]

    In: ICCV

    Wan, R., Shi, B., Duan, L.Y., Tan, A.H., Kot, A.C.: Benchmarking Single-Image Reflection Removal Algorithms. In: ICCV. pp. 3922–3930 (2017) 6

    work page 2017

  39. [39]

    In: AAAI

    Wang, J., Chan, K.C.K., Loy, C.C.: Exploring CLIP for assessing the look and feel of images. In: AAAI. pp. 2555–2563 (2023) 10

    work page 2023

  40. [40]

    International Journal of Computer Vision 132(12), 5929–5949 (2024) 10, 11, 12 DiffUR3 17

    Wang, J., Yue, Z., Zhou, S., Chan, K.C., Loy, C.C.: Exploiting Diffusion Prior for Real-World Image Super-Resolution. International Journal of Computer Vision 132(12), 5929–5949 (2024) 10, 11, 12 DiffUR3 17

    work page 2024

  41. [41]

    IEEE Transactions on Pattern Analysis and Machine Intelligence45(11), 12978–12995 (2023) 3

    Xiao, J., Fu, X., Liu, A., Wu, F., Zha, Z.J.: Image De-Raining Transformer. IEEE Transactions on Pattern Analysis and Machine Intelligence45(11), 12978–12995 (2023) 3

    work page 2023

  42. [42]

    IEEE Transactions on Pattern Analysis and Machine Intelli- gence38(9), 1721–1733 (2016) 1

    You, S., Tan, R.T., Kawakami, R., Ikeuchi, K.: Adherent Raindrop Detection and Removal in Video. IEEE Transactions on Pattern Analysis and Machine Intelli- gence38(9), 1721–1733 (2016) 1

    work page 2016

  43. [43]

    In: ICCV

    Zhang, L., Rao, A., Agrawala, M.: Adding conditional control to text-to-image diffusion models. In: ICCV. pp. 3836–3847 (2023) 9

    work page 2023

  44. [44]

    In: CVPR

    Zhang, R., Isola, P., Efros, A.A., Shechtman, E., Wang, O.: The unreasonable effectiveness of deep features as a perceptual metric. In: CVPR. pp. 586–595 (2018) 10

    work page 2018

  45. [45]

    In: CVPR

    Zhao, H., Li, M., Hu, Q., Guo, X.: Reversible decoupling network for single image reflection removal. In: CVPR. pp. 26430–26439 (2025) 2, 3, 12, 13

    work page 2025

  46. [46]

    In: CVPR

    Zheng, D., Wu, X.M., Yang, S., Zhang, J., Hu, J.F., Zheng, W.s.: Selective hour- glass mapping for universal image restoration based on diffusion model. In: CVPR. pp. 25445–25455 (2024) 12, 13

    work page 2024

  47. [47]

    Zheng, Q., Chen, J., Lu, Z., Shi, B., Jiang, X., Yap, K.H., Duan, L.Y., Kot, A.C.: What does plate glass reveal about camera calibration? In: CVPR. pp. 3022–3032 (2020) 1

    work page 2020

  48. [48]

    In: NeurIPS

    Zhou, S., Chan, K.C., Li, C., Loy, C.C.: Towards Robust Blind Face Restoration with Codebook Lookup Transformer. In: NeurIPS. vol. 35, pp. 30599–30611 (2022) 11

    work page 2022

  49. [49]

    IEEE Transactions on Geoscience and Remote Sensing63, 1–15 (2025) 2

    Zhu,C.,Deng,S.,Song,X.,Li,Y.,Wang,Q.:MambaCollaborativeImplicitNeural Representation for Hyperspectral and Multispectral Remote Sensing Image Fusion. IEEE Transactions on Geoscience and Remote Sensing63, 1–15 (2025) 2

    work page 2025

  50. [50]

    Information Fusion123, 103261 (2025) 2

    Zhu, C., Song, X., Li, Y., Deng, S., Zhang, T.: A spatial-frequency dual-domain implicitguidancemethodforhyperspectralandmultispectralremotesensingimage fusion based on Kolmogorov–Arnold Network. Information Fusion123, 103261 (2025) 2

    work page 2025

  51. [51]

    In: CVPR

    Zhu, Y., Fu, X., Jiang, P.T., Zhang, H., Sun, Q., Chen, J., Zha, Z.J., Li, B.: Revisiting Single Image Reflection Removal In the Wild. In: CVPR. pp. 25468– 25478 (2024) 2, 4

    work page 2024