Pith. sign in

REVIEW 4 major objections 5 minor 48 references

Freqformer: Image-Demoir\'eing Transformer via Efficient Frequency Decomposition

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read A dual-branch Transformer that splits moiré into high-frequency texture and low-frequency color can beat larger demoiréing networks on FHDMi and UHDM.

desk verdict A novel, compact demoiréing transformer with credible SOTA results, undermined by an undocumented scale-robustness claim and a training/inference contradiction. read the letter →

arxiv 2505.19120 v1 pith:6OTWOZFA submitted 2025-05-25 cs.CV

classification cs.CV
keywords imagedemoiréingfrequencydecompositiondual-branchtransformerlow-frequencycolordistortionhigh-frequencytexturechannelattentionrestorationFHDMi
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

Freqformer argues that moiré artifacts split cleanly by frequency: high-frequency components carry localized texture corruption, low-frequency components carry smooth color distortions. The paper builds a dual-branch Transformer that trains on these two components separately, using random crops for the high-frequency branch and heavily downsampled images for the low-frequency branch, then fuses the branches with a learnable Frequency Composition Transform. On the FHDMi and UHDM benchmarks, Freqformer reports the best PSNR, SSIM, and LPIPS on FHDMi and best SSIM/LPIPS on UHDM, with 6.065M parameters, smaller than most compared models. The paper's point is that explicit frequency separation, not a monolithic mapper, is what makes efficient high-resolution demoiréing feasible.

What carries the argument

The load-bearing object is the recursive frequency decomposition of Eq. (2): starting with $I_l^0 = I_m$, each level applies a dilated 3x3 convolution with the binomial kernel $k$ to get $I_l^i$, and defines the high-frequency residual as $I_h^i = I_h^{i-1} + I_l^{i-1} - I_l^i$. After $L$ levels, $I_l$ is the smooth color component and $I_h$ the texture component, both at full resolution. This decomposition does the work of making the two branches trainable with different strategies, and it feeds the learnable FCT, which combines the branch features before the final PixelShuffle reconstruction.

What would settle it

Take a diverse set of FHDMi and UHDM moiré images, apply the paper's L-level decomposition, downsample the low-frequency component to 0.1x, resize it back, and measure reconstruction PSNR against the original low-frequency component. If typical values fall well below roughly 50 dB, or if images with strong low-frequency scene texture break the scale-robustness, the resize-based training strategy degrades and the dual-branch design loses its justification.

Watch

Extended reading notes

Core claim

The paper claims that the right decomposition of a moiré image is into a low-frequency image that keeps color information and is robust to downsampling, and a high-frequency residual that keeps texture and is spatially local. This decomposition, built from L-level dilated convolutions with a fixed binomial kernel, separates moiré into components with different training needs: crop-based training for the high-frequency branch and resize-based training for the low-frequency branch. A learnable FCT then fuses the two branches in feature space instead of adding pixel outputs, which avoids accumulating reconstruction errors. With this design and a lightweight Spatial-Aware Channel Attention module, the paper reports results on FHDMi and UHDM that beat prior demoiréing and general restoration models, including much larger ones, at 6.065M parameters.

Load-bearing premise

The method assumes the low-frequency component can be downsampled to about one-tenth its size and resized back with roughly 50 dB PSNR, so training on resized low-frequency images is equivalent to training on full-resolution inputs; this is supported only by a one-sentence experiment without protocol details.

Editorial extensions

If this is right

  • If the decomposition claim is right, demoiréing can be solved by two smaller specialized networks rather than one large holistic network.
  • Resize-based low-frequency training means the color-correction branch can run at low resolution during training without hurting final quality, cutting memory and compute on 4K inputs.
  • The learnable FCT removes the need for a fixed inverse transform, so frequency fusion can be tuned to the data instead of being tied to Haar or other predefined wavelets.
  • The branch-level results in the paper (low branch 29.90 PSNR, high branch 29.20) suggest each branch is independently strong, so the gains are not just from the fusion module.

Reading between the lines

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

  • The scale-robust low-frequency assumption, if it generalizes, implies that any global color correction task (white balance, tint removal, cross-ISP color shifts) could be trained on downsampled images and then applied at full resolution with the same network.
  • The decomposition may transfer to other screen-capture artifacts such as flicker banding or subpixel aliasing, since those also have a smooth color component and a local texture component.
  • A testable extension would be to measure how far the resize factor can be pushed (0.05x, 0.02x) before the roughly 50 dB reconstruction PSNR of the low-frequency branch degrades, giving a direct compute-versus-quality trade-off for 8K inputs.
  • The paper compares against the 2-level Haar transform and reports a large drop (test PSNR 13.32), but it does not sweep the kernel or level count; varying those could reveal how much of the gain comes from this specific decomposition design.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes Freqformer, a dual-branch Transformer for image demoiréing that first decomposes the input into high-frequency and low-frequency components via a recursive, dilated low-pass convolution, then processes each branch with a tailored training strategy (crop for high frequency, resize for low frequency), and finally fuses the branches with a learnable Frequency Composition Transform (FCT). The authors report state-of-the-art results on the FHDMi and UHDM benchmarks with a compact 6.065M-parameter model, and support the design with ablations on the frequency decomposition, the FCT, the resize strategy, and the proposed SA-CA module.

Significance. If the results hold, Freqformer is a meaningful contribution to single-image demoiréing: it combines a frequency-decomposition design with an efficient dual-branch transformer and demonstrates competitive or superior performance against both dedicated demoiréing methods and recent general restoration models at a lower parameter count. The paper is strengthened by a public code release, comparisons on two public high-resolution benchmarks, and a coherent set of ablations that isolate the contributions of the main components. However, the central efficiency claim relies on a scale-robustness premise for the low-frequency branch that is asserted with minimal experimental detail, and the reported UHDM margins over the second-best method are small, so the SOTA claim needs stronger evidence and a clarified inference protocol.

major comments (4)
  1. [Sec. 3.2, bullet 2] The claim that the low-frequency component can be downsampled to 0.1x and reconstructed at ~50 dB PSNR is load-bearing for the resize-based training strategy and the low-scale inference, but the supporting experiment is not described: no decomposition level L, dataset, interpolation method, subset size, or variance is reported. Please provide the full protocol and results, including the value of L used, or alternatively reframe this premise as an empirical observation supported by the branch-level results in Table 3 rather than as a general property of the decomposition.
  2. [Sec. 3, first paragraph vs. Sec. 3.2, bullet 4] The test-time protocol is contradictory: Sec. 3 states that during testing both branches operate on the full-resolution image, while Sec. 3.2 bullet 4 states that the low-frequency branch uses the same small-scale input during both training and inference. Please specify the exact inference pipeline, particularly the input resolution of the low branch, how its output is upsampled, and how the upsampled features are aligned with the high-branch features in the FCT before the final reconstruction. This is also needed to interpret the reported 2.49 TFLOPS for 4K inference and the UHDM results.
  3. [Table 1] No error bars, repeated runs, or statistical significance tests are reported. On UHDM, Freqformer's advantage over ESDNet-L is small (SSIM 0.8018 vs. 0.7985; LPIPS 0.2434 vs. 0.2454), so single-run numbers do not establish that the ranking is stable. Please report at least three runs with mean and standard deviation, or a paired comparison, for the UHDM experiments.
  4. [Sec. 3.1, Eq. (2)] The number of decomposition levels L is never specified in the paper, although it controls the cutoff frequency of the low-frequency component and therefore directly affects the validity of the 0.1x resizing claim and the reproducibility of the method. Please state the value of L in Sec. 3.1 and list it in the implementation details.
minor comments (5)
  1. [Sec. 3.2] There are several typos in this section, including 'awesome benifits' and 'reals' (for 'reveals'), which should be corrected.
  2. [Sec. 2.1] The phrase 'Ultr-High-Definition' is missing a hyphen and should read 'Ultra-High-Definition'.
  3. [Sec. 3.1] The expression 'P ostF usion' appears with artificial spacing; this formatting should be fixed to 'PostFusion' throughout.
  4. [Table 4] The row labeled 'Freqformer-L (e2e)' is confusing because no 'Freqformer-L' variant is defined; please clarify whether this is an end-to-end version of Freqformer without frequency decomposition or a different variant.
  5. [Fig. 3] Some notation in Figure 3 (e.g., the projected features with prime symbols and the subscripts) is difficult to read at the printed size; consider enlarging or simplifying the notation in the figure.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: reported results are measured on held-out test data and the frequency-decomposition property, though under-supported, is an empirical claim rather than a definitional reduction.

full rationale

Freqformer's central claims are benchmark comparisons and an architecture; they are not derived from the model's equations in a way that presupposes the outcome. The frequency decomposition (Eq. 2) defines I_l and I_h as recursive convolution/difference, so I_l + I_h = I_m by construction, but the paper never predicts the final demoiréing quality from that identity; it trains on FHDMi/UHDM and reports held-out PSNR/SSIM/LPIPS (Tables 1-2). The resize-based training strategy rests on the empirical scale-robustness assertion in Sec. 3.2 (reconstruction PSNR ~50 dB for I_l), which is under-documented (no decomposition level, dataset, or protocol) and load-bearing for the UHDM margin, but this is a missing-support / reproducibility weakness, not circularity: the claim is an independent measurement, not a fitted parameter renamed as a prediction. The only overlapping-author citations ([39] for channel attention, [48] as a comparison baseline) are not load-bearing for the SOTA claim. The internal inconsistency between Sec. 3 ('In the testing phase, both branches operate on the full-resolution image') and Sec. 3.2 bullet 4 ('same small-scale input during both training and inference') is an ambiguity, not a circular reduction. No step in the paper reduces by construction to its own inputs.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

The central claim depends on hand-chosen design constants (kernel k, level L, loss weights, training scales) and on three domain assumptions about how moiré energy separates in frequency. No new physical entities are introduced.

free parameters (4)
  • Low-pass kernel k = [[1/16,1/8,1/16],[1/8,1/4,1/8],[1/16,1/8,1/16]]
    Hand-designed smoothing kernel in Eq. (1). The quality of the frequency separation, and therefore the whole dual-branch design, depends on it, but it is chosen rather than learned or derived.
  • Decomposition levels L = not reported
    The recursive decomposition in Eq. (2) runs for L levels, but L is never given in the implementation details. It controls the low-frequency/high-frequency split and must be guessed for reproduction.
  • Loss weights lambda1, lambda2 = lambda1 = 1, lambda2 = 0.1
    Set 'for simplicity' in Sec. 4. The balance between L1 and perceptual loss affects the final PSNR/SSIM/LPIPS trade-off and is a hand-chosen constant.
  • Training scales = FHDMi: crop 512, resize 512; UHDM: crop 768, resize 512
    The crop and resize sizes determine what the two branches see during training. They are selected by the authors and are load-bearing for the crop/resize strategy claim.
assumptions (3)
  • domain assumption Moiré patterns separate into high-frequency textures and low-frequency color distortions.
    The entire dual-branch design rests on this decomposition of moiré artifacts. The paper supports it with Fig. 2 visualizations and qualitative descriptions, not with a quantitative analysis of energy separation.
  • domain assumption The low-frequency component is scale-robust and can be downsampled to ~0.1x and resized back with ~50 dB reconstruction PSNR.
    This is load-bearing for the resize-based low-frequency branch. It is stated in Sec. 3.2 bullet 2 with a one-sentence, protocol-free experiment, so the reader cannot verify the magnitude of the loss.
  • domain assumption The high-frequency component has strong spatial locality, so random cropping is sufficient for training.
    Stated in Sec. 3.2 bullet 1. No experiment isolates locality; the crop-based training is justified mainly by the final performance of the whole system.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Freqformer: Image-Demoir\'eing Transformer via Efficient Frequency Decomposition." pith.science (2026). https://pith.science/paper/6OTWOZFA

@misc{pith2026250519120,
  author       = {Pith},
  title        = {Pith review of: Freqformer: Image-Demoir\'eing Transformer via Efficient Frequency Decomposition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6OTWOZFA}},
  note         = {Machine review of arXiv:2505.19120}
}
read the original abstract

Image demoir\'eing remains a challenging task due to the complex interplay between texture corruption and color distortions caused by moir\'e patterns. Existing methods, especially those relying on direct image-to-image restoration, often fail to disentangle these intertwined artifacts effectively. While wavelet-based frequency-aware approaches offer a promising direction, their potential remains underexplored. In this paper, we present Freqformer, a Transformer-based framework specifically designed for image demoir\'eing through targeted frequency separation. Our method performs an effective frequency decomposition that explicitly splits moir\'e patterns into high-frequency spatially-localized textures and low-frequency scale-robust color distortions, which are then handled by a dual-branch architecture tailored to their distinct characteristics. We further propose a learnable Frequency Composition Transform (FCT) module to adaptively fuse the frequency-specific outputs, enabling consistent and high-fidelity reconstruction. To better aggregate the spatial dependencies and the inter-channel complementary information, we introduce a Spatial-Aware Channel Attention (SA-CA) module that refines moir\'e-sensitive regions without incurring high computational cost. Extensive experiments on various demoir\'eing benchmarks demonstrate that Freqformer achieves state-of-the-art performance with a compact model size. The code is publicly available at https://github.com/xyLiu339/Freqformer.

Figures

Figures reproduced from arXiv: 2505.19120 by the authors.

Figure 1
Figure 1. Frequency Decomposition. To overcome these limitations, we introduce a novel frequency decomposition to efficiently separate moiré textures and color distortions into high-frequency and low-frequency components, respectively ( [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Visualization of the frequency-decomposed images (zoom in for better analysis). The left [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The whole pipeline of our Freqformer. The figure shows the dual-branch learning with [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Left part is the detailed transformer architecture of the SA-CA module utilized in all the [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Visual comparisons of different baselines on FHDMi and UHDM datasets (please zoom in [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 44 canonical work pages

  1. [1]

    Coarse-to-fine disentangling demoiréing framework for recaptured screen images

    Ce Wang, Bin He, Shengsen Wu, Renjie Wan, Boxin Shi, and Ling-Yu Duan. Coarse-to-fine disentangling demoiréing framework for recaptured screen images. TPAMI, 2023

  2. [2]

    Fhde 2 net: Full high definition demoireing network

    Bin He, Ce Wang, Boxin Shi, and Ling-Yu Duan. Fhde 2 net: Full high definition demoireing network. In ECCV, 2020

  3. [3]

    Mop moire patterns using mopnet

    Bin He, Ce Wang, Boxin Shi, and Ling-Yu Duan. Mop moire patterns using mopnet. In ICCV, 2019

  4. [4]

    Image demoireing with learnable bandpass filters

    Bolun Zheng, Shanxin Yuan, Gregory Slabaugh, and Ales Leonardis. Image demoireing with learnable bandpass filters. In ICCV, 2020

  5. [5]

    Moiré photo restoration using multiresolution convolutional neural networks

    Yujing Sun, Yizhou Yu, and Wenping Wang. Moiré photo restoration using multiresolution convolutional neural networks. TIP, 2018

  6. [6]

    Aim 2019 challenge on image demoireing: Methods and results

    Shanxin Yuan, Radu Timofte, Gregory Slabaugh, Aleš Leonardis, Bolun Zheng, Xin Ye, Xiang Tian, Yaowu Chen, Xi Cheng, Zhenyong Fu, et al. Aim 2019 challenge on image demoireing: Methods and results. In ICCVW, 2019

  7. [7]

    Towards efficient and scale-robust ultra-high-definition image demoiréing

    Xin Yu, Peng Dai, Wenbo Li, Lan Ma, Jiajun Shen, Jia Li, and Xiaojuan Qi. Towards efficient and scale-robust ultra-high-definition image demoiréing. In ECCV, 2022

  8. [8]

    Image demoireing in raw and srgb domains

    Shuning Xu, Binbin Song, Xiangyu Chen, Xina Liu, and Jiantao Zhou. Image demoireing in raw and srgb domains. In ECCV, 2024

Show all 48 references
  1. [9]

    P-bic: Ultra-high-definition image moiré patterns removal via patch bilateral compensation

    Zeyu Xiao, Zhihe Lu, and Xinchao Wang. P-bic: Ultra-high-definition image moiré patterns removal via patch bilateral compensation. In ACM MM, 2024

  2. [10]

    Deep wavelet network with domain adaptation for single image demoireing

    Xiaotong Luo, Jiangtao Zhang, Ming Hong, Yanyun Qu, Yuan Xie, and Cuihua Li. Deep wavelet network with domain adaptation for single image demoireing. In CVPRW, 2020

  3. [11]

    Wavelet-based dual-branch network for image demoiréing

    Lin Liu, Jianzhuang Liu, Shanxin Yuan, Gregory Slabaugh, Aleš Leonardis, Wengang Zhou, and Qi Tian. Wavelet-based dual-branch network for image demoiréing. In ECCV, 2020

  4. [12]

    Multibranch wavelet-based network for image demoiréing

    Chia-Hung Yeh, Chen Lo, and Cheng-Han He. Multibranch wavelet-based network for image demoiréing. Sensors, 2024

  5. [13]

    Hardware-friendly descreening

    Hasib Siddiqui, Mireille Boutin, and Charles A Bouman. Hardware-friendly descreening. TIP, 2009

  6. [14]

    Suppression of moiré patterns via spectral analysis

    Denis N Sidorov and Anil Christopher Kokaram. Suppression of moiré patterns via spectral analysis. In VCIP, 2002

  7. [15]

    Moiré pattern removal from texture images via low-rank and sparse matrix decomposition

    Fanglei Liu, Jingyu Yang, and Huanjing Yue. Moiré pattern removal from texture images via low-rank and sparse matrix decomposition. In VCIP, 2015

  8. [16]

    Real-time image demoireing on mobile devices

    Yuxin Zhang, Mingbao Lin, Xunchao Li, Han Liu, Guozhi Wang, Fei Chao, Shuai Ren, Yafei Wen, Xiaoxin Chen, and Rongrong Ji. Real-time image demoireing on mobile devices. ICLR, 2023

  9. [17]

    Doing more with moiré pattern detection in digital photos

    Cong Yang, Zhenyu Yang, Yan Ke, Tao Chen, Marcin Grzegorzek, and John See. Doing more with moiré pattern detection in digital photos. TIP, 2023

  10. [18]

    Image demoiréing via multi-scale fusion networks with moiré data augmentation

    Yan-Tsung Peng, Chih-Hsiang Hou, You-Cheng Lee, Aiden J Yoon, Zihao Chen, Yi-Ting Lin, and Wei- Cheng Lien. Image demoiréing via multi-scale fusion networks with moiré data augmentation. IEEE Sens. J., 2024

  11. [19]

    Video demoireing with relation-based temporal consistency

    Peng Dai, Xin Yu, Lan Ma, Baoheng Zhang, Jia Li, Wenbo Li, Jiajun Shen, and Xiaojuan Qi. Video demoireing with relation-based temporal consistency. In CVPR, 2022

  12. [20]

    Recaptured raw screen image and video demoireing via channel and spatial modulations

    Yijia Cheng, Xin Liu, and Jingyu Yang. Recaptured raw screen image and video demoireing via channel and spatial modulations. NeurIPS, 2023

  13. [21]

    Direction-aware video demoireing with temporal-guided bilateral learning

    Shuning Xu, Binbin Song, Xiangyu Chen, and Jiantao Zhou. Direction-aware video demoireing with temporal-guided bilateral learning. In AAAI, 2024

  14. [22]

    Recaptured screen image demoiréing in raw domain

    Huanjing Yue, Yijia Cheng, Yan Mao, Cong Cao, and Jingyu Yang. Recaptured screen image demoiréing in raw domain. TMM, 25, 2022

  15. [23]

    Wavelet convolutional neural networks for texture classification

    Shin Fujieda, Kohei Takayama, and Toshiya Hachisuka. Wavelet convolutional neural networks for texture classification. arXiv preprint arXiv:1707.07394, 2017. 10

  16. [24]

    Wavelet integrated cnns for noise-robust image classification

    Qiufu Li, Linlin Shen, Sheng Guo, and Zhihui Lai. Wavelet integrated cnns for noise-robust image classification. In CVPR, 2020

  17. [25]

    Scaling the scattering transform: Deep hybrid networks

    Edouard Oyallon, Eugene Belilovsky, and Sergey Zagoruyko. Scaling the scattering transform: Deep hybrid networks. In ICCV, 2017

  18. [26]

    Wavelet pooling for convolutional neural networks

    Travis Williams and Robert Li. Wavelet pooling for convolutional neural networks. In ICLR, 2018

  19. [27]

    Faster neural networks straight from jpeg

    Lionel Gueguen, Alex Sergeev, Ben Kadlec, Rosanne Liu, and Jason Yosinski. Faster neural networks straight from jpeg. NeurIPS, 2018

  20. [28]

    Convolutional neural network feature reduction using wavelet transform

    A Levinskis. Convolutional neural network feature reduction using wavelet transform. ELEKTRON ELEKTROTECH, 2013

  21. [29]

    Attribute-aware face aging with wavelet-based generative adversarial networks

    Yunfan Liu, Qi Li, and Zhenan Sun. Attribute-aware face aging with wavelet-based generative adversarial networks. In CVPR, 2019

  22. [30]

    Wavelet-srnet: A wavelet-based cnn for multi-scale face super resolution

    Huaibo Huang, Ran He, Zhenan Sun, and Tieniu Tan. Wavelet-srnet: A wavelet-based cnn for multi-scale face super resolution. In ICCV, 2017

  23. [31]

    Multi-level wavelet-cnn for image restoration

    Pengju Liu, Hongzhi Zhang, Kai Zhang, Liang Lin, and Wangmeng Zuo. Multi-level wavelet-cnn for image restoration. In CVPRW, 2018

  24. [32]

    Photorealistic style transfer via wavelet transforms

    Jaejun Yoo, Youngjung Uh, Sanghyuk Chun, Byeongkyu Kang, and Jung-Woo Ha. Photorealistic style transfer via wavelet transforms. In ICCV, 2019

  25. [33]

    Residual dense network for image super-resolution

    Yulun Zhang, Yapeng Tian, Yu Kong, Bineng Zhong, and Yun Fu. Residual dense network for image super-resolution. In CVPR, 2018

  26. [34]

    Densely connected convolu- tional networks

    Gao Huang, Zhuang Liu, Kilian Q Weinberger, and Laurens van der Maaten. Densely connected convolu- tional networks. In CVPR, 2017

  27. [35]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In CVPR, 2016

  28. [36]

    Multi-scale context aggregation by dilated convolutions

    Fisher Yu and Vladlen Koltun. Multi-scale context aggregation by dilated convolutions. ICLR, 2015

  29. [37]

    Chan, and Chen Change Loy

    Jianyi Wang, Zongsheng Yue, Shangchen Zhou, Kelvin C.K. Chan, and Chen Change Loy. Exploiting diffusion prior for real-world image super-resolution. IJCV, 2024

  30. [38]

    Restormer: Efficient transformer for high-resolution image restoration

    Syed Waqas Zamir, Aditya Arora, Salman Khan, Munawar Hayat, Fahad Shahbaz Khan, and Ming-Hsuan Yang. Restormer: Efficient transformer for high-resolution image restoration. In CVPR, 2022

  31. [39]

    Dual aggregation transformer for image super-resolution

    Zheng Chen, Yulun Zhang, Jinjin Gu, Linghe Kong, Xiaokang Yang, and Fisher Yu. Dual aggregation transformer for image super-resolution. In ICCV, 2023

  32. [40]

    Very deep convolutional networks for large-scale image recogni- tion

    Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale image recogni- tion. ICLR, 2015

  33. [41]

    Image quality assessment: from error visibility to structural similarity

    Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Simoncelli. Image quality assessment: from error visibility to structural similarity. TIP, 2004

  34. [42]

    The unreasonable effectiveness of deep features as a perceptual metric

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

  35. [43]

    Adam: A method for stochastic optimization

    Diederik Kingma and Jimmy Ba. Adam: A method for stochastic optimization. In ICLR, 2015

  36. [44]

    Sgdr: Stochastic gradient descent with warm restarts

    Ilya Loshchilov and Frank Hutter. Sgdr: Stochastic gradient descent with warm restarts. In ICLR, 2017

  37. [45]

    Multi-scale dynamic feature encoding network for image demoiréing

    Xi Cheng, Zhenyong Fu, and Jian Yang. Multi-scale dynamic feature encoding network for image demoiréing. In ICCVW, 2019

  38. [46]

    One-step effective diffusion network for real-world image super-resolution

    Rongyuan Wu, Lingchen Sun, Zhiyuan Ma, and Lei Zhang. One-step effective diffusion network for real-world image super-resolution. NeurIPS, 2024

  39. [47]

    AdaIR: Adaptive all-in-one image restoration via frequency mining and modulation

    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 modulation. In ICLR, 2025

  40. [48]

    Complexity experts are task-discriminative learners for any image restoration

    Eduard Zamfir, Zongwei Wu, Nancy Mehta, Yuedong Tan, Danda Pani Paudel, Yulun Zhang, and Radu Timofte. Complexity experts are task-discriminative learners for any image restoration. In CVPR, 2025. 11

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.