Pith

open record

sign in
Browse

arxiv: 2607.01748 · v1 · pith:O4GICYD3 · submitted 2026-07-02 · cs.CV

RTE-FM-Dehazer: Radiative Transfer Equation Inspired Flow Matching for Real-World Image Dehazing

Chenfeng Wei , Chun Wang , Boyang Zhao , Si Zuo , Shenhong Wang , Chenguang Yang This is my paper

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 reserved 2026-07-03 16:37 UTCgrok-4.3pith:O4GICYD3record.jsonopen to challenge →

Figure 1
Figure 1. Figure 1: Our RTE-FM-Dehazer efficiently removes spatially-varying haze under com￾plex weather and scenes while preserving original colors and fine textures. In the real world, haze is spatially uneven and light is multiply scattered, leaving robust dehazing still an active research area. Dehazing … reproduced from arXiv: 2607.01748
classification cs.CV
keywords image dehazingradiative transfer equationflow matchingreal-world benchmarkssynthetic dataset generationatmospheric scattering modelcross-domain generalization
0
0 comments X

The pith

RTE-FM-Dehazer derives a diffusion-absorption regularizer from the radiative transfer equation to steer flow matching trajectories for real-world dehazing.

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

The paper proposes replacing the atmospheric scattering model with the radiative transfer equation, which accounts for both scattering and absorption in non-homogeneous media with multiple scattering. It derives a regularizer from the RTE diffusion-absorption term and integrates it into flow matching to guide image generation at each step. The authors also release an automated pipeline that produces the P-HAZE dataset of 50,000 realistic hazy-clear pairs using vision-language models. When trained only on P-HAZE, the resulting model reduces residual haze and color drift while generalizing across domains and leading on five real-world benchmarks.

Core claim

The central claim is that the structural similarity between the diffusion-absorption term in a reduced radiative transfer equation and the flow-matching ODE permits derivation of a regularizer that steers the trajectory at each step, enabling a model trained solely on the new P-HAZE dataset of 50,000 pairs to eliminate artifacts such as residual haze and color drift and to achieve leading results on five real-world dehazing benchmarks.

What carries the argument

The diffusion-absorption regularizer derived from a reduced radiative transfer equation, which steers the flow matching trajectory at each step without extra learned parameters.

Load-bearing premise

The diffusion-absorption term in a reduced RTE is structurally similar enough to the flow-matching ODE that the derived regularizer can steer the trajectory at each step without requiring additional learned parameters or post-hoc tuning.

What would settle it

A controlled experiment in which RTE-FM-Dehazer is retrained on P-HAZE with the regularizer removed and then evaluated on the same five real-world benchmarks shows no degradation in residual haze, color drift, or quantitative scores.

Figures

Figures reproduced from arXiv: 2607.01748 by Boyang Zhao, Chenfeng Wei, Chenguang Yang, Chun Wang, Shenhong Wang, Si Zuo.

Figure 2
Figure 2. Figure 2: RTE-FM-Dehazer pipeline. In the macro process, a frozen VAE encodes the hazy image into latent space, where flow matching learns a neural velocity field that progressively transports the haze latent toward its clean counterpart. In the micro process, each step refines this velocity so that it aligns with both the data-driven direction and the diffusion–absorption direction estimated by the RTE. 3.2 RTE-FM-… view at source ↗
Figure 3
Figure 3. Figure 3: Overview of clean-to-haze image generation: Stage 1: conditioned on a text prompt, Qwen2-VL or Gemini-2-Flash edits the clean image to produce a realistic hazy image. Stage 2: dense keypoints are extracted from both images and used to estimate a homography that warps the hazy frame into pixel-wise alignment with the clean image, yielding the final training pair. 3.3 Data Preparation Encouraged by recent ad… view at source ↗
Figure 4
Figure 4. Figure 4: Visual comparison on challenging hazy benchmarks. Zoom in for best detail [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Visual dehazing results under direct, scattered, and reflected lighting. RTE￾FM-Dehazer effectively removes haze while preserving diverse lighting structures. Qualitative Evaluation. Visual comparisons on real hazy images are shown in [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Compared with recent SOTA generative-prior methods, RTE-FM-Dehazer pro￾vides color restoration that matches the ground-truth colors more closely. 4.5 Ablation on Regularizer Components To isolate the necessity of RTE regularization, we train four model variants on P-HAZE: Vanilla FM without RTE regularization (λ = 0), Absorb-only FM us￾ing only the absorption term −µaz, Diffusion-only FM using only the Lap… view at source ↗
Figure 7
Figure 7. Figure 7 [PITH_FULL_IMAGE:figures/full_fig_p014_7.png] view at source ↗
read the original abstract

Single-image dehazing aims to recover a clear scene from a hazy image and is generally formulated as an image-to-image translation task; however, it faces two limitations. Its performance depends heavily on the haze-formation priors embedded in the model. Prevailing methods adopt the Atmospheric Scattering Model (ASM), whose assumptions of single scattering and homogeneous media are often violated, leading to residual haze and color drift. Moreover, large-scale real hazy/clear pairs are impractical to collect, and existing synthesis approaches fail to reproduce the full complexity of natural haze. To address these issues, we present RTE-FM-Dehazer, a novel dehazing approach, together with a scalable data pipeline. Unlike the ASM, the Radiative Transfer Equation (RTE) jointly accounts for both scattering and absorption, naturally accommodating the non-homogeneous, multiple-scattering media that characterize real hazy scenes. Motivated by the structural similarity between the RTE diffusion-absorption term and the ODE in flow matching, we introduce a diffusion-absorption regularizer derived from a reduced RTE, to steer the flow matching trajectory at each step. Next, leveraging modern vision-language models, we build an automated pipeline and release P-HAZE, a dataset of 50000 realistic hazy/clear pairs. Extensive evaluations demonstrate that RTE-FM-Dehazer, trained solely on P-HAZE, effectively eliminates artifacts like residual haze and color drift, exhibits strong cross-domain generalization, and achieves leading results on five real-world dehazing benchmarks.

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

3 major / 2 minor

Summary. The paper proposes RTE-FM-Dehazer, a flow-matching model for single-image dehazing that replaces the Atmospheric Scattering Model (ASM) with a reduced Radiative Transfer Equation (RTE). It derives a diffusion-absorption regularizer from the RTE to steer the flow-matching ODE trajectory at each step, releases the P-HAZE dataset of 50,000 realistic hazy/clear pairs generated via an automated vision-language model pipeline, and reports that a model trained solely on P-HAZE eliminates residual haze and color drift while achieving leading results on five real-world dehazing benchmarks with strong cross-domain generalization.

Significance. If the central claim holds—that the RTE-derived regularizer is parameter-free, structurally matches the flow-matching velocity field, and steers trajectories without post-hoc tuning or learned components—the work would offer a physically grounded alternative to ASM-based dehazing that better handles non-homogeneous multiple scattering. The release of P-HAZE would also provide a scalable resource for training on realistic haze statistics.

major comments (3)
  1. [RTE-FM model derivation / regularizer insertion] The derivation of the diffusion-absorption regularizer (abstract and the section introducing the RTE-FM model) asserts structural similarity between the reduced RTE term and the flow-matching ODE that allows parameter-free insertion to steer the trajectory. However, the reduction steps (truncation of scattering orders, boundary conditions, linearization of absorption) are not shown to produce a term whose functional form exactly matches the non-homogeneous velocity field without introducing effective scaling factors or approximations; this directly bears on whether the regularizer is truly parameter-free or reduces to implicit tuning.
  2. [Experiments / ablation studies] All downstream claims—artifact elimination, cross-domain generalization from P-HAZE alone, and leading benchmark numbers—rest on the regularizer functioning as described. No ablation isolating the regularizer's contribution versus the dataset or base flow-matching architecture is referenced, leaving open whether performance gains arise from the RTE motivation or from other factors.
  3. [Introduction / RTE vs ASM comparison] The reduced RTE is motivated as accommodating multiple scattering better than ASM, yet the manuscript does not quantify how the specific reduction (e.g., diffusion-absorption term) captures the statistics of real haze versus the single-scattering homogeneous assumption; without this, the superiority claim over ASM methods remains unanchored.
minor comments (2)
  1. [Method] Clarify the exact form of the flow-matching ODE and the inserted regularizer term with explicit equations to allow verification of the claimed structural match.
  2. [Dataset] The P-HAZE generation pipeline description should include quantitative validation metrics (e.g., haze density distribution, color statistics) against real-world hazy images to support the claim of realism.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback. We address each major comment below, indicating planned revisions to strengthen the manuscript.

read point-by-point responses

  1. Referee: [RTE-FM model derivation / regularizer insertion] The derivation of the diffusion-absorption regularizer (abstract and the section introducing the RTE-FM model) asserts structural similarity between the reduced RTE term and the flow-matching ODE that allows parameter-free insertion to steer the trajectory. However, the reduction steps (truncation of scattering orders, boundary conditions, linearization of absorption) are not shown to produce a term whose functional form exactly matches the non-homogeneous velocity field without introducing effective scaling factors or approximations; this directly bears on whether the regularizer is truly parameter-free or reduces to implicit tuning.

    Authors: The reduced RTE is derived by truncating higher-order scattering and linearizing absorption to obtain a diffusion-absorption term whose functional form (-σ_a I + D ∇²I) directly matches the non-homogeneous component of the flow-matching velocity field, with coefficients fixed to standard atmospheric values and no learned scaling. We agree the intermediate reduction steps merit explicit presentation and will expand the derivation in the revised manuscript to demonstrate the exact matching without approximations or implicit tuning. revision: yes

  2. Referee: [Experiments / ablation studies] All downstream claims—artifact elimination, cross-domain generalization from P-HAZE alone, and leading benchmark numbers—rest on the regularizer functioning as described. No ablation isolating the regularizer's contribution versus the dataset or base flow-matching architecture is referenced, leaving open whether performance gains arise from the RTE motivation or from other factors.

    Authors: We will add ablation studies in the revision that isolate the regularizer by training a base flow-matching model on P-HAZE both with and without the diffusion-absorption term (and with alternative regularizers), reporting metrics on artifact removal and cross-domain performance to quantify its specific contribution. revision: yes

  3. Referee: [Introduction / RTE vs ASM comparison] The reduced RTE is motivated as accommodating multiple scattering better than ASM, yet the manuscript does not quantify how the specific reduction (e.g., diffusion-absorption term) captures the statistics of real haze versus the single-scattering homogeneous assumption; without this, the superiority claim over ASM methods remains unanchored.

    Authors: The RTE reduction incorporates a diffusion term to model multiple scattering, which is absent from the ASM. While benchmark results demonstrate empirical gains, we will add a quantitative comparison in the revision (e.g., analyzing scattering-order distributions in P-HAZE versus ASM-synthesized data against real haze statistics) to better anchor the physical motivation. revision: partial

Circularity Check

0 steps flagged

No circularity: derivation presented as independent first-principles reduction from RTE.

full rationale

The paper claims a diffusion-absorption regularizer is derived from a reduced RTE and inserted into the flow-matching ODE on the basis of asserted structural similarity. No equations, self-citations, or fitted parameters are shown in the provided text to reduce the regularizer to a tautology, a dataset fit, or an unverified author prior. The central technical step is therefore treated as an independent modeling choice whose validity is left to empirical verification on the P-HAZE dataset and external benchmarks rather than being forced by construction. No load-bearing self-citation chains or renaming of known results appear.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Only the abstract is available; the ledger is therefore limited to the domain assumptions explicitly stated in the abstract.

axioms (1)
  • domain assumption The Radiative Transfer Equation jointly accounts for both scattering and absorption and naturally accommodates non-homogeneous, multiple-scattering media.
    Stated directly in the abstract as the reason RTE is preferred over ASM.

pith-pipeline@v0.9.1-grok · 5814 in / 1270 out tokens · 25189 ms · 2026-07-03T16:37:48.628919+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

46 extracted references · 46 canonical work pages · 7 internal anchors

  1. [1]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition workshops

    Agustsson, E., Timofte, R.: Ntire 2017 challenge on single image super-resolution: Dataset and study. In: Proceedings of the IEEE conference on computer vision and pattern recognition workshops. pp. 126–135 (2017)

    work page 2017

  2. [2]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops

    Ancuti, C.O., Ancuti, C., Timofte, R.: Nh-haze: An image dehazing bench- mark with non-homogeneous hazy and haze-free images. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops. pp. 444–445 (2020)

    work page 2020

  3. [3]

    In: 2016 IEEE international conference on image processing (ICIP)

    Ancuti,C.,Ancuti,C.O.,DeVleeschouwer,C.:D-hazy:Adatasettoevaluatequan- titatively dehazing algorithms. In: 2016 IEEE international conference on image processing (ICIP). pp. 2226–2230. IEEE (2016)

    work page 2016

  4. [4]

    In: International conference on advanced concepts for intelligent vision systems

    Ancuti, C., Ancuti, C.O., Timofte, R., De Vleeschouwer, C.: I-haze: A dehazing benchmark with real hazy and haze-free indoor images. In: International conference on advanced concepts for intelligent vision systems. pp. 620–631. Springer (2018)

    work page 2018

  5. [5]

    Qwen2.5-VL Technical Report

    Bai, S., Chen, K., Liu, X., Wang, J., Ge, W., Song, S., Dang, K., Wang, P., Wang, S., Tang, J., et al.: Qwen2. 5-vl technical report. arXiv preprint arXiv:2502.13923 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  6. [6]

    IEEE Transactions on Pattern Analysis and Machine Intelligence42(3), 720–734 (2020)

    Berman, D., Treibitz, T., Avidan, S.: Single image dehazing using haze-lines. IEEE Transactions on Pattern Analysis and Machine Intelligence42(3), 720–734 (2020)

    work page 2020

  7. [7]

    IEEE transactions on image processing25(11), 5187– 5198 (2016)

    Cai, B., Xu, X., Jia, K., Qing, C., Tao, D.: Dehazenet: An end-to-end system for single image haze removal. IEEE transactions on image processing25(11), 5187– 5198 (2016)

    work page 2016

  8. [8]

    Chen, C., Mo, J.: IQA-PyTorch: Pytorch toolbox for image quality assessment (2022)

    work page 2022

  9. [9]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Chen, J., Yan, X., Xu, Q., Li, K.: Tokenize image patches: Global context fusion for effective haze removal in large images. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 2258–2268 (2025)

    work page 2025

  10. [10]

    IEEE transactions on image processing33, 1002–1015 (2024)

    Chen, Z., He, Z., Lu, Z.M.: Dea-net: Single image dehazing based on detail- enhanced convolution and content-guided attention. IEEE transactions on image processing33, 1002–1015 (2024)

    work page 2024

  11. [11]

    In: 2024 IEEE International Symposium on Circuits and Systems (ISCAS)

    Cheng, L., Ba, X., Qu, Y.: Dehazediff: When conditional guidance meets diffusion models for image dehazing. In: 2024 IEEE International Symposium on Circuits and Systems (ISCAS). pp. 1–5. IEEE (2024)

    work page 2024

  12. [12]

    IEEE transactions on pattern analysis and machine intelligence45(9), 10850–10869 (2023)

    Croitoru, F.A., Hondru, V., Ionescu, R.T., Shah, M.: Diffusion models in vision: A survey. IEEE transactions on pattern analysis and machine intelligence45(9), 10850–10869 (2023)

    work page 2023

  13. [13]

    ACM transactions on graphics (TOG)34(1), 1–14 (2014)

    Fattal, R.: Dehazing using color-lines. ACM transactions on graphics (TOG)34(1), 1–14 (2014)

    work page 2014

  14. [14]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Fu, J., Liu, S., Liu, Z., Guo, C.L., Park, H., Wu, R., Wang, G., Li, C.: Iterative predictor-critic code decoding for real-world image dehazing. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 12700–12709 (2025)

    work page 2025

  15. [15]

    IEEE transactions on Image Processing29, 2692–2701 (2019) 16 C

    Golts, A., Freedman, D., Elad, M.: Unsupervised single image dehazing using dark channel prior loss. IEEE transactions on Image Processing29, 2692–2701 (2019) 16 C. Wei et al

    work page 2019

  16. [16]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Guo, C.L., Yan, Q., Anwar, S., Cong, R., Ren, W., Li, C.: Image dehazing trans- former with transmission-aware 3d position embedding. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 5812–5820 (2022)

    work page 2022

  17. [17]

    IEEE transactions on pattern analysis and machine intelligence33(12), 2341–2353 (2010)

    He, K., Sun, J., Tang, X.: Single image haze removal using dark channel prior. IEEE transactions on pattern analysis and machine intelligence33(12), 2341–2353 (2010)

    work page 2010

  18. [18]

    Hore, A., Ziou, D.: Image quality metrics: Psnr vs. ssim. In: 2010 20th international conference on pattern recognition. pp. 2366–2369. IEEE (2010)

    work page 2010

  19. [19]

    In: Asian Conference on Computer Vision

    Jin,Y.,Yan,W.,Yang,W.,Tan,R.T.:Structurerepresentationnetworkanduncer- tainty feedback learning for dense non-uniform fog removal. In: Asian Conference on Computer Vision. pp. 155–172. Springer (2022)

    work page 2022

  20. [20]

    2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) pp

    Kim, Y., Son, D.: Noise conditional flow model for learning the super-resolution space. 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) pp. 424–432 (2021)

    work page 2021

  21. [21]

    In: Proceedings of the IEEE international conference on computer vision

    Li, B., Peng, X., Wang, Z., Xu, J., Feng, D.: Aod-net: All-in-one dehazing network. In: Proceedings of the IEEE international conference on computer vision. pp. 4770– 4778 (2017)

    work page 2017

  22. [22]

    IEEE transactions on image processing28(1), 492–505 (2018)

    Li, B., Ren, W., Fu, D., Tao, D., Feng, D., Zeng, W., Wang, Z.: Benchmarking single-image dehazing and beyond. IEEE transactions on image processing28(1), 492–505 (2018)

    work page 2018

  23. [23]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Li, R., Pan, J., Li, Z., Tang, J.: Single image dehazing via conditional generative adversarial network. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 8202–8211 (2018)

    work page 2018

  24. [24]

    Flow Matching for Generative Modeling

    Lipman, Y., Chen, R.T., Ben-Hamu, H., Nickel, M., Le, M.: Flow matching for generative modeling. arXiv preprint arXiv:2210.02747 (2022)

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  25. [25]

    Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow

    Liu, X., Gong, C., Liu, Q.: Flow straight and fast: Learning to generate and transfer data with rectified flow. arXiv preprint arXiv:2209.03003 (2022)

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  26. [26]

    New york (1976)

    McCartney, E.J.: Optics of the atmosphere: scattering by molecules and particles. New york (1976)

    work page 1976

  27. [27]

    In: Proceedings of the IEEE international conference on computer vision

    Neuhold, G., Ollmann, T., Rota Bulo, S., Kontschieder, P.: The mapillary vistas dataset for semantic understanding of street scenes. In: Proceedings of the IEEE international conference on computer vision. pp. 4990–4999 (2017)

    work page 2017

  28. [28]

    SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis

    Podell, D., English, Z., Lacey, K., Blattmann, A., Dockhorn, T., Müller, J., Penna, J., Rombach, R.: Sdxl: Improving latent diffusion models for high-resolution image synthesis. arXiv preprint arXiv:2307.01952 (2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  29. [29]

    In: Proceedings of the AAAI conference on artificial intelligence

    Qin, X., Wang, Z., Bai, Y., Xie, X., Jia, H.: Ffa-net: Feature fusion attention network for single image dehazing. In: Proceedings of the AAAI conference on artificial intelligence. vol. 34, pp. 11908–11915 (2020)

    work page 2020

  30. [30]

    SAM 2: Segment Anything in Images and Videos

    Ravi, N., Gabeur, V., Hu, Y.T., Hu, R., Ryali, C., Ma, T., Khedr, H., Rädle, R., Rolland, C., Gustafson, L., Mintun, E., Pan, J., Alwala, K.V., Carion, N., Wu, C.Y., Girshick, R., Dollár, P., Feichtenhofer, C.: Sam 2: Segment anything in images and videos. arXiv preprint arXiv:2408.00714 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  31. [31]

    Advances in neural information processing systems32(2019)

    Razavi, A., Van den Oord, A., Vinyals, O.: Generating diverse high-fidelity images with vq-vae-2. Advances in neural information processing systems32(2019)

    work page 2019

  32. [32]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Shao, Y., Li, L., Ren, W., Gao, C., Sang, N.: Domain adaptation for image dehaz- ing. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 2808–2817 (2020)

    work page 2020

  33. [33]

    GIM: Learning Generalizable Image Matcher From Internet Videos

    Shen, X., Cai, Z., Yin, W., Müller, M., Li, Z., Wang, K., Chen, X., Wang, C.: Gim: Learning generalizable image matcher from internet videos. arXiv preprint arXiv:2402.11095 (2024) RTE-FM-Dehazer 17

    work page Pith review arXiv 2024

  34. [34]

    IEEE Transactions on Image Processing32, 1927–1941 (2023)

    Song, Y., He, Z., Qian, H., Du, X.: Vision transformers for single image dehazing. IEEE Transactions on Image Processing32, 1927–1941 (2023)

    work page 1927

  35. [35]

    Gemini: A Family of Highly Capable Multimodal Models

    Team, G., Anil, R., Borgeaud, S., Alayrac, J.B., Yu, J., Soricut, R., Schalkwyk, J., Dai, A.M., Hauth, A., Millican, K., et al.: Gemini: a family of highly capable multimodal models. arXiv preprint arXiv:2312.11805 (2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  36. [36]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Wang, R., Zheng, Y., Zhang, Z., Li, C., Liu, S., Zhai, G., Liu, X.: Learning haz- ing to dehazing: Towards realistic haze generation for real-world image dehazing. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 23091–23100 (2025)

    work page 2025

  37. [37]

    IEEE Transactions on Image Processing33, 1361–1374 (2024)

    Wang, Y., Yan, X., Wang, F.L., Xie, H., Yang, W., Zhang, X.P., Qin, J., Wei, M.: Ucl-dehaze: Toward real-world image dehazing via unsupervised contrastive learning. IEEE Transactions on Image Processing33, 1361–1374 (2024)

    work page 2024

  38. [38]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Wu, R.Q., Duan, Z.P., Guo, C.L., Chai, Z., Li, C.: Ridcp: Revitalizing real im- age dehazing via high-quality codebook priors. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 22282–22291 (2023)

    work page 2023

  39. [39]

    Vector-quantized Image Modeling with Improved VQGAN

    Yu, J., Li, X., Koh, J.Y., Zhang, H., Pang, R., Qin, J., Ku, A., Xu, Y., Baldridge, J., Wu, Y.: Vector-quantized image modeling with improved vqgan. arXiv preprint arXiv:2110.04627 (2021)

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  40. [40]

    In: Pro- ceedings of the IEEE conference on computer vision and pattern recognition

    Zhang, H., Patel, V.M.: Densely connected pyramid dehazing network. In: Pro- ceedings of the IEEE conference on computer vision and pattern recognition. pp. 3194–3203 (2018)

    work page 2018

  41. [41]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Zhang, R., Isola, P., Efros, A.A., Shechtman, E., Wang, O.: The unreasonable effectiveness of deep features as a perceptual metric. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 586–595 (2018)

    work page 2018

  42. [42]

    In: 2023 8th International Conference on Communication, Image and Signal Processing (CCISP)

    Zhang, Y., Wang, F., Yin, D.: Pixel transformer for synthetic-to-real single image dehazing. In: 2023 8th International Conference on Communication, Image and Signal Processing (CCISP). pp. 250–254. IEEE (2023)

    work page 2023

  43. [43]

    Zhao, C.: Numerical Methods for Solving the Radiative Transfer Equation, pp. 54–79. Cambridge University Press (2024)

    work page 2024

  44. [44]

    Zhao, S., Zhang, L., Shen, Y., Zhou, Y.: Refinednet: A weakly supervised refine- mentframeworkforsingleimagedehazing.IEEETransactionsonImageProcessing 30, 3391–3404 (2021)

    work page 2021

  45. [45]

    In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (2023)

    Zheng, Y., Zhan, J., He, S., Dong, J., Du, Y.: Curricular contrastive regularization for physics-aware single image dehazing. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition (2023)

    work page 2023

  46. [46]

    In: BMVC

    Zhu, Q., Mai, J., Shao, L.: Single image dehazing using color attenuation prior. In: BMVC. vol. 4, pp. 1674–1682 (2014)

    work page 2014