arxiv: 2604.03800 · v1 · submitted 2026-04-04 · 💻 cs.CV

HistoFusionNet: Histogram-Guided Fusion and Frequency-Adaptive Refinement for Nighttime Image Dehazing

Mohammad Heydari , Wei Dong , Shahram Shirani , Jun Chen , Han Zhou This is my paper

Pith reviewed 2026-05-13 17:20 UTC · model grok-4.3

classification 💻 cs.CV

keywords nighttime image dehazinghistogram transformerfrequency-adaptive refinementimage restorationlow-light visionmulti-scale encoder-decoderNTIRE challenge

0 comments

The pith

HistoFusionNet groups features by dynamic range in transformer blocks and refines them with frequency cues to restore nighttime hazy images.

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

The paper introduces HistoFusionNet as a unified architecture for nighttime image dehazing that must handle the simultaneous effects of haze, glow, uneven illumination, color shifts, and noise. It uses a multi-scale encoder-decoder with histogram transformer blocks that group features according to their intensity ranges, allowing long-range dependencies among similarly degraded regions to be modeled directly. A separate frequency-aware branch then balances low-frequency scene structure against high-frequency detail recovery. The resulting framework is presented as effective for the heterogeneous degradations typical of real nighttime scenes, with top performance reported on the NTIRE 2026 benchmark.

Core claim

Histogram transformer blocks that group features by dynamic-range characteristics enable better aggregation of degraded regions, while a frequency-adaptive refinement branch exploits complementary low- and high-frequency information to suppress artifacts and recover details, together forming a single model suited to the mixture of degradations found in nighttime hazy images.

What carries the argument

Histogram transformer blocks that group features by dynamic-range characteristics, paired with a frequency-aware refinement branch inside a multi-scale encoder-decoder.

If this is right

Nighttime dehazing no longer needs separate modules for each degradation type.
Scene structures and local textures recover more reliably by explicit low- and high-frequency balancing.
The same architecture can process varied real-world nighttime conditions without domain-specific tuning.
Competitive benchmark rankings follow directly from the unified handling of mixed degradations.

Where Pith is reading between the lines

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

Similar histogram grouping could be tested on other intensity-varying tasks such as low-light enhancement.
The frequency branch might reduce flickering when applied to video sequences of moving scenes.
Autonomous-vehicle perception at night could gain from feeding these restored frames into downstream detectors.
The approach suggests that intensity-range clustering is a general lever for any restoration problem dominated by non-uniform lighting.

Load-bearing premise

Grouping features solely by intensity range and splitting refinement by frequency bands is enough to capture the interactions among haze, glow, uneven light, color distortion, and noise without extra priors.

What would settle it

A held-out set of real nighttime hazy images containing strong localized glow and sensor noise where the restored outputs show more visible artifacts or lost detail than a standard multi-scale baseline would falsify the claim that the histogram and frequency components provide the necessary advantage.

Figures

Figures reproduced from arXiv: 2604.03800 by Han Zhou, Jun Chen, Mohammad Heydari, Shahram Shirani, Wei Dong.

**Figure 2.** Figure 2: Overall architecture of HistoFusionNet. Our dehazing network adopts a U-shaped design with a DCNv4-based main branch and an auxiliary frequency-aware branch. Histogram transformer blocks are inserted at the bottleneck to perform dynamic-range aware global aggregation, while a lightweight frequency-adaptive refinement module is employed to enhance color fidelity and recover fine details. 2. Related Work Nig… view at source ↗

**Figure 3.** Figure 3: Visual comparisons on NH-HAZE dataset. Compared to other models, our method exhibits higher color fidelity and effective [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

**Figure 4.** Figure 4: Visual experiment results on NH-HAZE2 dataset. Obviously, our method demonstrates superior performance on color preserva [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗

**Figure 5.** Figure 5: Qualitative comparison on the Dense-Haze dataset. Our method produces clearer structures, better color fidelity, and more [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗

**Figure 6.** Figure 6: Our results on the validation set of the NTIRE 2026 Nighttime Image Dehazing Challenge, demonstrating strong dehazing [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗

read the original abstract

Nighttime image dehazing remains a challenging low-level vision problem due to the joint presence of haze, glow, non-uniform illumination, color distortion, and sensor noise, which often invalidate assumptions commonly used in daytime dehazing. To address these challenges, we propose HistoFusionNet, a transformer-enhanced architecture tailored for nighttime image dehazing by combining histogram-guided representation learning with frequency-adaptive feature refinement. Built upon a multi-scale encoder-decoder backbone, our method introduces histogram transformer blocks that model long-range dependencies by grouping features according to their dynamic-range characteristics, enabling more effective aggregation of similarly degraded regions under complex nighttime lighting. To further improve restoration fidelity, we incorporate a frequency-aware refinement branch that adaptively exploits complementary low- and high-frequency cues, helping recover scene structures, suppress artifacts, and enhance local details. This design yields a unified framework that is particularly well suited to the heterogeneous degradations encountered in real nighttime hazy scenes. Extensive experiments and highly competitive performance of our method on the NTIRE 2026 Nighttime Image Dehazing Challenge benchmark demonstrate the effectiveness of the proposed method. Our team ranked 1st among 22 participating teams, highlighting the robustness and competitive performance of HistoFusionNet. The code is available at: https://github.com/heydarimo/Night-Time-Dehazing

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.

Desk Editor's Note private letter to a colleague

HistoFusionNet won the NTIRE 2026 nighttime dehazing challenge with a transformer that groups by histogram/dynamic range plus a frequency refinement branch, but the write-up does not isolate whether those pieces drove the result.

read the letter

The headline result is that this architecture took first place among 22 teams on the NTIRE 2026 Nighttime Image Dehazing Challenge benchmark. The authors built it on a multi-scale encoder-decoder, added histogram transformer blocks that group features by dynamic-range characteristics, and attached a frequency-aware refinement branch to handle low- and high-frequency cues separately. Code is released, which is helpful for anyone who wants to reproduce or extend the work. Those two additions are the concrete new elements; the rest follows standard transformer and multi-scale patterns already used in restoration tasks. The approach makes sense for nighttime scenes where haze, glow, uneven lighting, and noise appear together, and the top ranking shows it works at least as well as the other entries. The main gap is the missing controls. There are no ablations that remove the histogram grouping or the frequency branch to measure the drop, and no direct comparison against a plain multi-scale transformer baseline trained the same way. Without those, the performance gain could come from the backbone, data handling, or training schedule rather than the proposed components. The paper would be tighter if it showed that the new blocks are load-bearing. This is useful for researchers who need a ready-to-run method for nighttime dehazing or similar low-level restoration problems. A reader who wants a competitive baseline with public code will find it worth looking at. I would send it to peer review; the empirical result and released implementation are enough to justify referee time, even if the authors need to add the missing ablation analysis.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces HistoFusionNet, a transformer-enhanced multi-scale encoder-decoder architecture for nighttime image dehazing. It incorporates histogram transformer blocks that group features by dynamic-range characteristics to model long-range dependencies among similarly degraded regions, together with a frequency-aware refinement branch that adaptively combines low- and high-frequency cues. The method is evaluated on the NTIRE 2026 Nighttime Image Dehazing Challenge benchmark, where the authors report a first-place ranking among 22 participating teams.

Significance. If the reported ranking is supported by rigorous controls, the work supplies a unified framework explicitly designed for the joint degradations (haze, glow, non-uniform illumination, color distortion, sensor noise) that invalidate standard daytime dehazing priors. The open-sourced code at the provided GitHub link strengthens reproducibility and allows the community to verify the claimed robustness on real nighttime scenes.

major comments (2)

[Experiments and Results] The central claim that the histogram-guided and frequency-adaptive components specifically address heterogeneous nighttime degradations rests on the NTIRE 2026 ranking, yet the manuscript provides no ablation that removes either the histogram transformer blocks or the frequency-aware refinement branch while keeping the backbone, training schedule, and data augmentation fixed. Without these controls, the performance gain cannot be isolated from the encoder-decoder architecture itself.
[Experiments and Results] Table 1 (or equivalent quantitative comparison table) reports only the final ranking and aggregate metrics; it does not include per-component PSNR/SSIM deltas or comparisons against a plain multi-scale transformer baseline trained under identical conditions. This omission leaves open the possibility that the reported advantage is attributable to training details rather than the proposed modules.

minor comments (2)

[Method] The abstract states that the histogram transformer blocks 'enable more effective aggregation of similarly degraded regions' but does not define the dynamic-range grouping operation mathematically; a short equation or pseudocode in §3 would clarify the implementation.
[Experiments and Results] Figure 4 (qualitative results) would benefit from zoomed insets highlighting recovery of fine structures and suppression of glow artifacts, as the current scale makes it difficult to assess local fidelity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback and positive assessment of the work's significance and reproducibility. We agree that the current experiments would benefit from explicit ablations to isolate module contributions and will incorporate these in the revision.

read point-by-point responses

Referee: [Experiments and Results] The central claim that the histogram-guided and frequency-adaptive components specifically address heterogeneous nighttime degradations rests on the NTIRE 2026 ranking, yet the manuscript provides no ablation that removes either the histogram transformer blocks or the frequency-aware refinement branch while keeping the backbone, training schedule, and data augmentation fixed. Without these controls, the performance gain cannot be isolated from the encoder-decoder architecture itself.

Authors: We agree that dedicated ablations are necessary to isolate the contributions. In the revised manuscript we will add experiments that remove the histogram transformer blocks and the frequency-aware refinement branch individually while freezing the multi-scale encoder-decoder backbone, training schedule, optimizer, and all data augmentations. We will report the resulting PSNR/SSIM drops on the NTIRE 2026 test set to quantify the specific gains attributable to each module. revision: yes
Referee: [Experiments and Results] Table 1 (or equivalent quantitative comparison table) reports only the final ranking and aggregate metrics; it does not include per-component PSNR/SSIM deltas or comparisons against a plain multi-scale transformer baseline trained under identical conditions. This omission leaves open the possibility that the reported advantage is attributable to training details rather than the proposed modules.

Authors: We acknowledge the omission. The revised manuscript will expand the quantitative table (or add a dedicated ablation table) to include (i) per-component PSNR/SSIM deltas for each removed module and (ii) direct comparison against a plain multi-scale transformer baseline trained from scratch under identical conditions, hyperparameters, and data. This will rule out training-detail confounds and clearly attribute performance differences to the proposed histogram-guided and frequency-adaptive components. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical architecture validated on external benchmark

full rationale

The paper proposes HistoFusionNet as a transformer-based architecture with histogram-guided blocks and frequency-adaptive refinement for nighttime dehazing. No mathematical derivation chain, fitted parameters, or first-principles predictions are present in the provided text. The central claim of suitability for heterogeneous degradations is supported solely by empirical ranking (1st on NTIRE 2026 benchmark), which is an external outcome rather than a quantity defined in terms of the model's own components or self-citations. No self-definitional reductions, uniqueness theorems, or ansatzes imported via prior work appear; the design choices are presented as motivated engineering decisions validated experimentally.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only access provides no information on free parameters, axioms, or invented entities; a full manuscript would be required to populate this ledger.

pith-pipeline@v0.9.0 · 5551 in / 1144 out tokens · 44532 ms · 2026-05-13T17:20:33.709975+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

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

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

histogram transformer blocks that model long-range dependencies by grouping features according to their dynamic-range characteristics
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

frequency-aware refinement branch that adaptively exploits complementary low- and high-frequency cues

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

61 extracted references · 61 canonical work pages

[1]

Ancuti, Cosmin Ancuti, Radu Timofte, Luc Van Gool, Lei Zhang, and Ming-Hsuan Yang

Codruta O. Ancuti, Cosmin Ancuti, Radu Timofte, Luc Van Gool, Lei Zhang, and Ming-Hsuan Yang. Ntire 2019 image dehazing challenge report. InProceedings of the IEEE Con- ference on Computer Vision and Pattern Recognition Work- shops (CVPR Workshops), 2019

work page 2019
[2]

Ancuti, Cosmin Ancuti, and Radu Timofte

Codruta O. Ancuti, Cosmin Ancuti, and Radu Timofte. Nh- haze: An image dehazing benchmark with nonhomogeneous hazy and haze-free images. InProceedings of the IEEE Con- ference on Computer Vision and Pattern Recognition Work- shops (CVPR Workshops), 2020

work page 2020
[3]

Ancuti, Cosmin Ancuti, Florin-Alexandru Vasluianu, Radu Timofte, et al

Codruta O. Ancuti, Cosmin Ancuti, Florin-Alexandru Vasluianu, Radu Timofte, et al. Ntire 2020 challenge on non- homogeneous dehazing. InProceedings of the IEEE Con- ference on Computer Vision and Pattern Recognition Work- shops (CVPR Workshops), 2020

work page 2020
[4]

Ancuti, Cosmin Ancuti amd Florin- Alexandru Vasluianu, and Radu Timofte

Codruta O. Ancuti, Cosmin Ancuti amd Florin- Alexandru Vasluianu, and Radu Timofte. Ntire 2021 nonhomogeneous dehazing challenge report. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPR Workshops), 2021

work page 2021
[5]

Ancuti, Cosmin Ancuti, Florin-Alexandru Vasluianu, and Radu Timofte

Codruta O. Ancuti, Cosmin Ancuti, Florin-Alexandru Vasluianu, and Radu Timofte. Ntire 2021 nonhomogeneous dehazing challenge report. InProceedings of the IEEE Con- ference on Computer Vision and Pattern Recognition Work- shops (CVPR Workshops), 2021

work page 2021
[6]

Ancuti, Cosmin Ancuti, Florin-Alexandru Vasluianu, Radu Timofte, et al

Codruta O. Ancuti, Cosmin Ancuti, Florin-Alexandru Vasluianu, Radu Timofte, et al. Ntire 2023 hr nonhomo- geneous dehazing challenge report. InProceedings of the IEEE Conference on Computer Vision and Pattern Recogni- tion Workshops (CVPR Workshops), 2023

work page 2023
[7]

Ancuti et al

Codruta O. Ancuti et al. Ntire 2026 nighttime image de- hazing challenge report. InProceedings of the IEEE Con- ference on Computer Vision and Pattern Recognition Work- shops (CVPR Workshops), 2026

work page 2026
[8]

Ancuti et al

Codruta O. Ancuti et al. A benchmark dataset for realistic night-time image dehzing. InProceedings of the IEEE Con- ference on Computer Vision and Pattern Recognition Work- shops (CVPR Workshops), 2026

work page 2026
[9]

Non-local im- age dehazing

Dana Berman, Tali Treibitz, and Shai Avidan. Non-local im- age dehazing. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016

work page 2016
[10]

Dehazenet: An end-to-end system for single image haze removal.IEEE Transactions on Image Process- ing (TIP), 2016

Bolun Cai, Xiangmin Xu, Kui Jia, Chunmei Qing, and Dacheng Tao. Dehazenet: An end-to-end system for single image haze removal.IEEE Transactions on Image Process- ing (TIP), 2016

work page 2016
[11]

Psd: Principled synthetic-to-real dehazing guided by physical priors

Zhitong Chen, Yang Wang, Yang Yang, Ding Liu, Junjie Cao, Yuemei Ren, Ming-Ming Cheng, Qian Wang, and Bain- ing Guo. Psd: Principled synthetic-to-real dehazing guided by physical priors. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2021

work page 2021
[12]

Ancuti, Florin-Alexandru Vasluianu, Radu Timofte, and et al

Cosmin Ancuti Codruta O. Ancuti, Florin-Alexandru Vasluianu, Radu Timofte, and et al. Ntire 2024 dense and non-homogeneous dehazing challenge report. InProceed- ings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPR Workshops), 2024

work page 2024
[13]

A semi-supervised nighttime dehazing baseline with spatial-frequency aware and realistic brightness constraint

Xiaofeng Cong, Jie Gui, Jing Zhang, Junming Hou, and Hao Shen. A semi-supervised nighttime dehazing baseline with spatial-frequency aware and realistic brightness constraint. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2024

work page 2024
[14]

Selective frequency network for image restoration

Yuning Cui, Yi Tao, Zhenshan Bing, Wenqi Ren, Xinwei Gao, Xiaochun Cao, Kai Huang, and Alois Knoll. Selective frequency network for image restoration. InProceedings of the International Conference on Learning Representations (ICLR), 2023

work page 2023
[15]

Adair: Adaptive all-in-one image restoration via frequency mining and mod- ulation

Yuning Cui, Syed Waqas Zamir, Salman Khan, Alois Knoll, Mubarak Shah, and Fahad Shahbaz Khan. Adair: Adaptive all-in-one image restoration via frequency mining and mod- ulation. InProceedings of the International Conference on Learning Representations (ICLR), 2025

work page 2025
[16]

Deformable convolutional networks

Jifeng Dai, Haozhi Qi, Yuwen Xiong, Yi Li, Guodong Zhang, Han Hu, and Yichen Wei. Deformable convolutional networks. InProceedings of the IEEE International Confer- ence on Computer Vision (ICCV), 2017

work page 2017
[17]

Multi-scale boosted de- hazing network with dense feature fusion

Hang Dong, Jinshan Pan, Lei Xiang, Zhe Hu, Xinyi Zhang, Fei Wang, and Ming-Hsuan Yang. Multi-scale boosted de- hazing network with dense feature fusion. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2020

work page 2020
[18]

Shadowrefiner: Towards mask-free shadow removal via fast fourier trans- former

Wei Dong, Han Zhou, Yuqiong Tian, Jingke Sun, Xiao- hong Liu, Guangtao Zhai, and Jun Chen. Shadowrefiner: Towards mask-free shadow removal via fast fourier trans- former. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPR Work- shops), 2024

work page 2024
[19]

Dehazedct: Towards effective non- homogeneous dehazing via deformable convolutional trans- former

Wei Dong, Han Zhou, Ruiyi Wang, Xiaohong Liu, Guang- tao Zhai, and Jun Chen. Dehazedct: Towards effective non- homogeneous dehazing via deformable convolutional trans- former. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPR Work- shops), 2024

work page 2024
[20]

Ecmamba: Consolidating selective state space model with retinex guidance for efficient multiple exposure correc- tion.Advances in Neural Information Processing Systems, 2024

Wei Dong, Han Zhou, Yulun Zhang, Xiaohong Liu, and Jun Chen. Ecmamba: Consolidating selective state space model with retinex guidance for efficient multiple exposure correc- tion.Advances in Neural Information Processing Systems, 2024

work page 2024
[21]

To- wards scale-aware low-light enhancement via structure- guided transformer design

Wei Dong, Yan Min, Han Zhou, and Jun Chen. To- wards scale-aware low-light enhancement via structure- guided transformer design. InProceedings of the IEEE Con- ference on Computer Vision and Pattern Recognition Work- shops (CVPR Workshops), 2025

work page 2025
[22]

Retinex-guided histogram transformer for mask-free shadow removal

Wei Dong, Han Zhou, Seyed Amirreza Mousavi, and Jun Chen. Retinex-guided histogram transformer for mask-free shadow removal. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPR Workshops), 2025

work page 2025
[23]

Towards adap- tive adverse weather removal via semantic and low-level vi- sual perceptual priors.Machine Learning and Knowledge Extraction, 2026

Wei Dong, Han Zhou, Terry Ji, and Jun Chen. Towards adap- tive adverse weather removal via semantic and low-level vi- sual perceptual priors.Machine Learning and Knowledge Extraction, 2026

work page 2026
[24]

Zero- reference joint low-light enhancement and deblurring via vi- sual autoregressive modeling with vlm-derived modulation

Wei Dong, Han Zhou, Junwei Lin, and Jun Chen. Zero- reference joint low-light enhancement and deblurring via vi- sual autoregressive modeling with vlm-derived modulation. 9 InProceedings of the AAAI Conference on Artificial Intelli- gence (AAAI), 2026

work page 2026
[25]

Single image dehazing.ACM Transactions on Graphics (TOG), 27(3):72:1–72:9, 2008

Raanan Fattal. Single image dehazing.ACM Transactions on Graphics (TOG), 27(3):72:1–72:9, 2008

work page 2008
[26]

Image dehazing transformer with transmission-aware 3d position embedding

Chun-Le Guo, Qixin Yan, Saeed Anwar, Runmin Cong, Wenqi Ren, and Chongyi Li. Image dehazing transformer with transmission-aware 3d position embedding. InProceed- ings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2022

work page 2022
[27]

Single image haze removal using dark channel prior

Kaiming He, Jian Sun, and Xiaoou Tang. Single image haze removal using dark channel prior. InProceedings of the IEEE Conference on Computer Vision and Pattern Recog- nition (CVPR), 2009

work page 2009
[28]

Focal frequency loss for image reconstruction and synthe- sis

Liming Jiang, Bo Dai, Wayne Wu, and Chen Change Loy. Focal frequency loss for image reconstruction and synthe- sis. InProceedings of the IEEE International Conference on Computer Vision (ICCV), 2021

work page 2021
[29]

When fast fourier transform meets transformer for image restoration

Xingyu Jiang, Xiuhui Zhang, Ning Gao, and Yue Deng. When fast fourier transform meets transformer for image restoration. InProceedings of the European Conference on Computer Vision (ECCV), 2024

work page 2024
[30]

Aod-net: All-in-one dehazing network

Boyi Li, Xiulian Peng, Zhangyang Wang, Jizheng Xu, and Dan Feng. Aod-net: All-in-one dehazing network. InPro- ceedings of the IEEE International Conference on Computer Vision (ICCV), 2017

work page 2017
[31]

Benchmarking single- image dehazing and beyond.IEEE Transactions on Image Processing (TIP), 28(1):492–505, 2019

Boyi Li, Wenqi Ren, Dengpan Fu, Dacheng Tao, Dan Feng, Wenjun Zeng, and Zhangyang Wang. Benchmarking single- image dehazing and beyond.IEEE Transactions on Image Processing (TIP), 28(1):492–505, 2019

work page 2019
[32]

Tan, and Michael S

Yu Li, Robby T. Tan, and Michael S. Brown. Nighttime haze removal with glow and multiple light colors. InProceedings of the IEEE International Conference on Computer Vision (ICCV), 2015

work page 2015
[33]

Swinir: Image restoration using swin transformer

Jingyun Liang, Jiezhang Cao, Guolei Sun, Kai Zhang, Luc Van Gool, and Radu Timofte. Swinir: Image restoration using swin transformer. InProceedings of the IEEE Inter- national Conference on Computer Vision Workshops (ICCV Workshops), 2021

work page 2021
[34]

Beibei Lin, Yeying Jin, Wending Yan, Wei Ye, Yuan Yuan, and Robby T. Tan. Nighthaze: Nighttime image dehazing via self-prior learning. InProceedings of the AAAI Conference on Artificial Intelligence (AAAI), 2025

work page 2025
[35]

Frequency domain-based diffusion model for unpaired image dehazing

Chengxu Liu, Lu Qi, Jinshan Pan, Xueming Qian, and Ming- Hsuan Yang. Frequency domain-based diffusion model for unpaired image dehazing. InProceedings of the IEEE Inter- national Conference on Computer Vision (ICCV), 2025

work page 2025
[36]

Trident dehazing network

Jing Liu, Yuan Xie Haiyan Wu, Yanyun Qu, and Lizhuang Ma. Trident dehazing network. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPR Workshops), 2020

work page 2020
[37]

Grid- dehazenet: Attention-based multi-scale network for image dehazing

Xiaohong Liu, Yongrui Ma, Zhihao Shi, and Jun Chen. Grid- dehazenet: Attention-based multi-scale network for image dehazing. InProceedings of the IEEE International Confer- ence on Computer Vision (ICCV), 2019

work page 2019
[38]

Nighttime image dehazing based on variational decom- position model

Yun Liu, Zhongsheng Yan, Aimin Wu, Tian Ye, and Yuche Li. Nighttime image dehazing based on variational decom- position model. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPR Workshops), 2022

work page 2022
[39]

Swin transformer: Hierarchical vision transformer using shifted windows

Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE International Conference on Com- puter Vision (ICCV), 2021

work page 2021
[40]

W. E. K. Middleton.Vision through the atmosphere. Univer- sity of Toronto Press, 1952

work page 1952
[41]

Ffa-net: Feature fusion attention network for single image dehazing

Xu Qin, Zhilin Wang, Yuanchao Bai, Xiaodong Xie, and Huizhu Jia. Ffa-net: Feature fusion attention network for single image dehazing. InProceedings of the AAAI Confer- ence on Artificial Intelligence (AAAI), 2020

work page 2020
[42]

Gated fusion net- work for single image dehazing

Wenqi Ren, Lin Ma, Jiawei Zhang, Jinshan Pan, Xiaochun Cao, Wei Liu, and Ming-Hsuan Yang. Gated fusion net- work for single image dehazing. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018

work page 2018
[43]

Spatial frequency modulation network for efficient image dehazing

Hao Shen, Henghui Ding, Yulun Zhang, Zhong-Qiu Zhao, and Xudong Jiang. Spatial frequency modulation network for efficient image dehazing. InIEEE Transactions on Image Processing (TIP), 2025

work page 2025
[44]

Vision transformers for single image dehazing.IEEE Transactions on Image Processing (TIP), 2023

Yuda Song, Zhuqing He, Hui Qian, and Xin Du. Vision transformers for single image dehazing.IEEE Transactions on Image Processing (TIP), 2023

work page 2023
[45]

Restoring images in adverse weather conditions via histogram transformer

Shangquan Sun, Wenqi Ren, Xinwei Gao, Rui Wang, and Xi- aochun Cao. Restoring images in adverse weather conditions via histogram transformer. InProceedings of the European Conference on Computer Vision (ECCV), 2024

work page 2024
[46]

Towards image ambi- ent lighting normalization

Florin-Alexandru Vasluianu, Tim Seizinger, Zongwei Wu, Rakesh Ranjan, and Radu Timofte. Towards image ambi- ent lighting normalization. InProceedings of the European Conference on Computer Vision (ECCV), 2024

work page 2024
[47]

Internim- age: Exploring large-scale vision foundation models with de- formable convolutions

Wenhai Wang, Jifeng Dai, Zhe Chen, Zhenhang Huang, Zhiqi Li, Xizhou Zhu, Xiaowei Hu, Tong Lu, Lewei Lu, Hongsheng Li, Xiaogang Wang, and Yu Qiao. Internim- age: Exploring large-scale vision foundation models with de- formable convolutions. InProceedings of the IEEE Confer- ence on Computer Vision and Pattern Recognition (CVPR), 2023

work page 2023
[48]

Chan, Ke Yu, Chao Dong, and Chen Change Loy

Xintao Wang, Kelvin C.K. Chan, Ke Yu, Chao Dong, and Chen Change Loy. Edvr: Video restoration with enhanced deformable convolutional networks. InProceedings of the IEEE Conference on Computer Vision and Pattern Recogni- tion Workshops (CVPR Workshops), 2019

work page 2019
[49]

Light field im- age super-resolution using deformable convolution

Yingqian Wang, Jungang Yang, Longguang Wang, Xinyi Ying, Tianhao Wu, Wei An, and Yulan Guo. Light field im- age super-resolution using deformable convolution. InIEEE Transactions on Image Processing (TIP), 2021

work page 2021
[50]

Uformer: A gen- eral u-shaped transformer for image restoration

Zhendong Wang, Xiaodong Cun, Jianmin Bao, Wengang Zhou, Jianzhuang Liu, and Houqiang Li. Uformer: A gen- eral u-shaped transformer for image restoration. InProceed- ings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2022

work page 2022
[51]

Con- trastive learning for compact single image dehazing

Haiyan Wu, Yanyun Qu, Shaohui Lin, Jian Zhou, Ruizhi Qiao, Zhizhong Zhang, Yuan Xie, and Lizhuang Ma. Con- trastive learning for compact single image dehazing. InPro- 10 ceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2021

work page 2021
[52]

Efficient de- formable convnets: Rethinking dynamic and sparse operator for vision applications

Yuwen Xiong, Zhiqi Li, Yuntao Chen, Feng Wang, Xizhou Zhu, Jiapeng Luo, Wenhai Wang, Tong Lu, Hongsheng Li, Yu Qiao, Lewei Lu, Jie Zhou, and Jifeng Dai. Efficient de- formable convnets: Rethinking dynamic and sparse operator for vision applications. InProceedings of the IEEE Confer- ence on Computer Vision and Pattern Recognition (CVPR), 2024

work page 2024
[53]

Wavelet u-net and the chromatic adaptation transform for single image dehazing

Hao-Hsiang Yang and Yanwei Fu. Wavelet u-net and the chromatic adaptation transform for single image dehazing. In Proceedings of the IEEE International Conference on Image Processing (ICIP), 2019

work page 2019
[54]

Frequency and spatial dual guidance for image dehazing

Hu Yu, Naishan Zheng, Man Zhou, Jie Huang, Zeyu Xiao, and Feng Zhao. Frequency and spatial dual guidance for image dehazing. InProceedings of the European Conference on Computer Vision (ECCV), 2022

work page 2022
[55]

Restormer: Efficient transformer for high-resolution image restoration

Syed Waqas Zamir, Aditya Arora, Salman Khan, Mu- nawar Hayat, Fahad Shahbaz Khan, and Ming-Hsuan Yang. Restormer: Efficient transformer for high-resolution image restoration. InProceedings of the IEEE Conference on Com- puter Vision and Pattern Recognition (CVPR), 2022

work page 2022
[56]

He Zhang and Vishal M. Patel. Densely connected pyramid dehazing network. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018

work page 2018
[57]

Break- ing through the haze: An advanced non-homogeneous de- hazing method based on fast fourier convolution and con- vnext

Han Zhou, Wei Dong, Yangyi Liu, and Jun Chen. Break- ing through the haze: An advanced non-homogeneous de- hazing method based on fast fourier convolution and con- vnext. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPR Work- shops), 2023

work page 2023
[58]

Glare: Low light image enhancement via generative latent feature based codebook retrieval

Han Zhou, Wei Dong, Xiaohong Liu, Shuaicheng Liu, Xiongkuo Min, Guangtao Zhai, and Jun Chen. Glare: Low light image enhancement via generative latent feature based codebook retrieval. InProceedings of the European Confer- ence on Computer Vision (ECCV), 2024

work page 2024
[59]

Lita-gs: Illumination- agnostic novel view synthesis via reference-free 3d gaus- sian splatting and physical priors

Han Zhou, Wei Dong, and Jun Chen. Lita-gs: Illumination- agnostic novel view synthesis via reference-free 3d gaus- sian splatting and physical priors. InProceedings of the IEEE Conference on Computer Vision and Pattern Recog- nition (CVPR), 2025

work page 2025
[60]

Low-light image enhancement via generative perceptual priors

Han Zhou, Wei Dong, Xiaohong Liu, Yulun Zhang, Guang- tao Zhai, and Jun Chen. Low-light image enhancement via generative perceptual priors. InProceedings of the AAAI Conference on Artificial Intelligence (AAAI), 2025

work page 2025
[61]

De- formable convnets v2: More deformable, better results

Xizhou Zhu, Stephen Lin Han Hu, and Jifeng Dai. De- formable convnets v2: More deformable, better results. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2019. 11

work page 2019