Pith. sign in

REVIEW 3 major objections 4 minor 101 references

Hierarchical Information Flow for Generalized Efficient Image Restoration

T0 review · 3 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The paper proposes Hi-IR, a transformer that replaces long-range self-attention with a three-level hierarchical information flow, and reports state-of-the-art results on seven image restoration tasks at lower parameter counts.

desk verdict Solid all-in-one restoration model; the experiments are broad and the scaling recipe is useful, but the Appendix C global-range claim is not supported. read the letter →

arxiv 2411.18588 v1 pith:NZYPS2EZ submitted 2024-11-27 cs.CV

Yawei Li , Bin Ren , Jingyun Liang , Rakesh Ranjan , Mengyuan Liu , Nicu Sebe , Ming-Hsuan Yang , Luca Benini This is my paper · ORCID
classification cs.CV
keywords imagerestorationhierarchicalinformationflowvisiontransformerself-attentionmodelscalingsuper-resolutiondenoisingdeblurring
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper argues that a restoration transformer does not need global self-attention or shifted windows: a fixed three-level hierarchy can push information from local patches to global context at window-attention cost. It builds Hi-IR, a model following this principle, and reports state-of-the-art accuracy on seven image restoration tasks with fewer parameters than comparable transformer baselines. The work also explains why larger restoration models train poorly, blaming initialization and gradient scale, and offers three fixes: learning-rate warmup, replacing dense 3×3 convolutions with bottlenecks, and dot-product attention. If the claims hold, one architecture can serve many degradations and restoration models can be scaled beyond the usual 10–20M parameter range.

What carries the argument

The central object is a hierarchical information tree realized in each Hi-IR layer: L1 self-attention inside p-by-p patches, L2 self-attention across s-by-s groups of dispersed patches obtained by a permutation operation, and L3 a convolutional feed-forward network of 1×1, 3×3, and 1×1 convolutions. The permutation step lets distant patches meet in one window without global attention. The per-layer time complexity is $O(BHW((5+2\gamma)C^2 + \frac{3}{2}(p^2+s^2)C))$, comparable to window attention, while two consecutive layers reach a $16P\times16P$ receptive field.

What would settle it

Train a parameter-matched shifted-window SwinIR variant on the same 4× super-resolution data and compare Urban100 PSNR against Hi-IR's reported 28.44 dB; if the control matches or exceeds it under identical training, the hierarchy is not the source of the gain.

Watch

Extended reading notes

Core claim

Hi-IR's central claim is that global information exchange in image restoration can be achieved by incrementally propagating information through a three-level tree, avoiding both quadratic global self-attention and the receptive-field limits of window attention. Level one applies self-attention inside small local patches; level two permutes dispersed patches into larger groups so each window receives cross-patch information; level three is a convolutional feed-forward network. The paper reports that this sequence reaches a global receptive field with space and time complexity comparable to window attention, and that a model built on it achieves the best or second-best PSNR/SSIM across super-resolution, denoising, JPEG artifact removal, motion deblurring, defocus deblurring, demosaicking, and adverse-weather restoration.

Load-bearing premise

The load-bearing premise is that a fixed three-level hierarchy is the right way to propagate information globally for all seven degradations, rather than simply having more layers or wider channels at the same parameter count.

Editorial extensions

If this is right

  • A single model architecture without global self-attention can reach top or runner-up accuracy across seven restoration tasks, so task-specific attention designs are not necessary on these benchmarks.
  • The scaling recipe—warmup, bottleneck convolutions, and dot-product attention—lets a restoration transformer grow from about 15M to roughly 57M parameters without the performance collapse previously seen, opening a path beyond the common 10–20M range.
  • Because the L2 permutation uses a bounded group size, the model limits how far apart attending pixels can be, which keeps computation predictable on large images.
  • Extending the tree depth beyond three levels improves accuracy further, making tree depth a usable tuning knob for the architecture.
  • A single Hi-IR checkpoint can handle multiple degradation levels in denoising and JPEG artifact removal, suggesting one model can replace per-degradation deployments.

Reading between the lines

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

  • Beyond the paper: the bounded L2 group size is a transferable inductive bias, so swapping window-attention blocks for L1–L2–L3 blocks in other high-resolution dense prediction tasks, such as semantic segmentation or depth estimation, is a direct test worth running at equal FLOPs.
  • Beyond the paper: the ablation study never compares against a same-parameter-capacity model with more layers or wider channels but no hierarchy, so part of the reported gain could be capacity rather than the information-flow mechanism; a parameter-matched shifted-window control would settle this.
  • Beyond the paper: the three scaling fixes are separable from the architecture, so applying warmup, bottleneck convolutions, and dot-product attention to other large transformer restoration baselines would show whether the recipe generalizes beyond Hi-IR.
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

3 major / 4 minor

Summary. The paper proposes Hi-IR, a transformer-based image restoration architecture built on a three-level hierarchical information flow: L1 self-attention within p×p patches, L2 attention over permuted groups of s² patches within P×P blocks (P = s·p), and L3 convolutional feed-forward processing. The authors argue that this hierarchy removes long-range self-attention, improves efficiency, and enables global information propagation at window-attention cost. They further propose scaling strategies (learning-rate warmup, replacing dense 3×3 convolutions with bottlenecks, and using dot-product attention) and validate the model on seven image restoration tasks: super-resolution, denoising, JPEG compression artifact removal, motion deblurring, defocus deblurring, demosaicking, and adverse-weather restoration. The central claims are that Hi-IR achieves state-of-the-art results in all seven tasks and that the hierarchical mechanism propagates information globally more efficiently than window attention.

Significance. If the central efficiency claim were correct, Hi-IR would be a valuable contribution: it would show that a transformer without global or shifted-window self-attention can reach global-range information flow at low cost, with strong results across many degradations. The experimental campaign is extensive, covering seven tasks, multiple datasets, two architecture families, and model-scaling analyses; the paper also reports several second-best results honestly rather than hiding them. The scaling analysis in Section 4, in particular, is a useful empirical study. However, the paper's main theoretical justification for global-range propagation is not supported as written, and the blanket state-of-the-art claim is stronger than the tables warrant. The contribution is therefore presently a strong empirical architecture with an unverified mechanism-level narrative.

major comments (3)
  1. [Appx. C, §3.2] The claim that "the maximum receptive field of two consecutive transformer layers is 16P" (Appx. C, after Eq. (8), and Table 19) is not supported by the mechanism described in §3.2. In the L2 step, the permutation groups tokens only within one P×P block, and there is no shift, random shuffle, or cross-block connection between aligned Hi-IR layers. A token in the columnar architecture therefore cannot receive evidence from outside its initial P×P block after any number of layers; the receptive field grows only by the O(1) kernel extent of the L3 convolution, not to 16P. Consequently, Table 19's comparison with "Window Attn. (8P×8P)", which reaches 16P via shifted windows, is not an equal-receptive-field comparison, and the statement in §3.1 that the proposed mechanism is "more efficient in propagating information to the global range" does not follow from the derivation. For the U-shape configuration used in most tasks, the encoder-decoder down-sampling supplies cross-block information flow independently of the L1/L2 attention, so the global-range claim should be restricted and re-derived.
  2. [§5.2, Tables 8, 12, 16] The abstract and introduction claim "state-of-the-art performance in seven common image restoration tasks", but several entries in the paper's own tables are second-best: in Tab. 8, Hi-IR-B and Hi-IR-L are below HAT on Set5 at ×2, ×3, and ×4 (e.g., 33.22 vs. 33.30 dB at ×4); in Tab. 12, Hi-IR-L is below GRL-B on HIDE (31.64 vs. 31.65 dB); and in Tab. 16, Hi-IR is below All-in-One on RainDrop (30.84 vs. 31.12 dB). The text in §5.2 acknowledges some of these cases, but the abstract and introduction should be qualified, for example to "competitive or state-of-the-art on most benchmarks and tasks", or the specific tasks for which the claim is intended should be stated explicitly.
  3. [§5.1, Tab. 6] The ablation in Tab. 6 varies the L1/L2 layout, the projection, the tree depth, and the model size simultaneously, but it does not include an equal-capacity control, such as a v3 or v4 model with more layers or wider channels at the same parameter count, or a plain window-attention model at the same capacity. Without such a control, the PSNR differences attributed to the hierarchical information flow could instead reflect the changed capacity or the modified training recipe, so the causal conclusion that the hierarchy itself is responsible for the gains is not established. Please add matched-capacity baselines or soften the causal claim.
minor comments (4)
  1. [§4.1] The sentence "none of the four methods improves the convergence and performance of the scaled models, indicating that they do work for the attention modules of the IR transformers" appears to contain a typo: based on Tab. 4, the intended meaning is likely "do not work".
  2. [§3.2] In the description of the second MSA, the query, key, and value projections are denoted Q^{l1}, K^{l1}, V^{l1}; they should presumably be indexed by l2 to match Y^{l2}.
  3. [Appx. D, Tab. 21] Tab. 21 is labeled "TreeIR" in its caption, while the method is called Hi-IR everywhere else; please make the naming consistent.
  4. [§5, Appx. A] The paper does not report error bars or statistical significance for the small performance gaps (e.g., 0.01 dB on HIDE, 0.02–0.04 dB on several SR settings), and no code release is mentioned; adding error bars or code would substantially strengthen the empirical claims.

Circularity Check

0 steps flagged · score 0.0 of 10

No construction-level circularity found; benchmark claims are external and self-citations are baseline comparisons, though Appx C's 16P receptive-field assertion is unsupported.

full rationale

The paper's central claims, namely that Hi-IR achieves state-of-the-art results on seven IR tasks and offers a better efficiency/accuracy trade-off, are supported by benchmark experiments and by complexity formulas in Sec. 5 and Appx. C. No quantity is fitted to a subset of the data and then reported as a prediction forced by that fit. The architecture choices, including the L1/L2 permutation attention, L3 convolutional FFN, dot-product attention, warm-up, and bottleneck convolutions, are validated on standard test sets, which is ordinary empirical practice rather than circularity. The most visible self-citation, GRL (Li et al., 2023a), is an overlapping-author baseline that Hi-IR is compared against, not a premise used to justify the method; the cited results appear only as baselines in Tabs. 7-13 and are externally reproducible. No uniqueness theorem or architectural premise is imported from the authors' prior work. The ablation in Tab. 6 does not include an equal-parameter depth/width control, so the marginal attribution of gains to the hierarchy itself is under-supported, but an omitted control is a validity concern, not a circular step. One load-bearing statement in Appx. C, "The maximum receptive field of two consecutive transformer layers is 16 P," is asserted without derivation and appears inconsistent with the aligned, non-overlapping P by P grouping described in Sec. 3.2; under that mechanism two layers should not see beyond one P by P block. This would undermine the global-range efficiency comparison with 8P-window attention, but an unsupported or incorrect derivation is not an equivalence-to-input, so under the operating rules it does not raise the circularity score.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The paper introduces no physical entities. Its free parameters are architectural hyperparameters and training choices selected through validation. The main assumptions are about the effectiveness of hierarchical information flow and the transferability of scaling tricks across tasks.

free parameters (5)
  • L1 patch size p = not reported in main text
    Controls the local attention window in the first level of the hierarchy. Chosen by hand; affects complexity and receptive field.
  • L2 permutation group size s = not reported in main text
    Controls the scope of cross-patch attention in the second level. Chosen by hand; appears in the complexity formulas.
  • Bottleneck reduction factor = 4
    The middle 3x3 convolution in the conv3 bottleneck block has channels reduced by a factor of 4. Selected based on the scaling experiments in Tab. 3.
  • Warmup iterations = 50k
    A warmup period of 50k iterations is chosen to stabilize large-model training, as described in Sec. 4.2.
  • Attention type = dot product
    Dot-product attention is selected because Tab. 5 shows it outperforms cosine similarity attention on SR benchmarks. This is a design choice, not a derived optimum.
assumptions (5)
  • domain assumption Progressive local-to-global information flow is necessary for efficient image restoration.
    Motivates the L1-L2-L3 hierarchy. The paper does not prove this; ablations only compare variants of the hierarchy.
  • domain assumption Replacing dense 3x3 convolutions with bottleneck or linear layers stabilizes large-model training.
    Sec. 4.3 gives a fan-in/fan-out heuristic based on Xavier/Kaiming initialization, not a proof. It is validated only on SR benchmarks.
  • domain assumption Warmup helps large-model convergence.
    Sec. 4.4 cites general warmup literature and empirical observations, but does not provide a rigorous derivation specific to image restoration.
  • domain assumption Dot-product attention has better gradient stability than cosine similarity attention.
    Sec. 4.5 derives the gradient expressions but the conclusion about instability relies on numerical observation in Fig. 4, which is not shown in detail.
  • standard math The complexity formulas in Appendix C correctly account for the proposed attention.
    The formulas are standard arithmetic for attention and feed-forward complexity, but they are not machine-checked and depend on the chosen p and s values.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Hierarchical Information Flow for Generalized Efficient Image Restoration." pith.science (2026). https://pith.science/paper/NZYPS2EZ

@misc{pith2026241118588,
  author       = {Pith},
  title        = {Pith review of: Hierarchical Information Flow for Generalized Efficient Image Restoration},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NZYPS2EZ}},
  note         = {Machine review of arXiv:2411.18588}
}
read the original abstract

While vision transformers show promise in numerous image restoration (IR) tasks, the challenge remains in efficiently generalizing and scaling up a model for multiple IR tasks. To strike a balance between efficiency and model capacity for a generalized transformer-based IR method, we propose a hierarchical information flow mechanism for image restoration, dubbed Hi-IR, which progressively propagates information among pixels in a bottom-up manner. Hi-IR constructs a hierarchical information tree representing the degraded image across three levels. Each level encapsulates different types of information, with higher levels encompassing broader objects and concepts and lower levels focusing on local details. Moreover, the hierarchical tree architecture removes long-range self-attention, improves the computational efficiency and memory utilization, thus preparing it for effective model scaling. Based on that, we explore model scaling to improve our method's capabilities, which is expected to positively impact IR in large-scale training settings. Extensive experimental results show that Hi-IR achieves state-of-the-art performance in seven common image restoration tasks, affirming its effectiveness and generalizability.

Figures

Figures reproduced from arXiv: 2411.18588 by the authors.

Figure 1
Figure 1. The proposed Hi-IR is notable for its efficiency and effectiveness (a)-(b), generalizability across [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Illustration of information flow principles. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Illustrations of: (a) The hierarchical information flow. (b) The proposed hierarchical information [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (13 more)
Figure 4
Figure 4. Figure 4: Comparsion of gradients between dot product and cosine similarity. optimizer to find a stable path in the loss landscape before applying larger updates. Warmup enables the model to adapt gradually, avoiding overshooting minima and ensuring smoother convergence. 4.5 Why…
Figure 5
Figure 5. Figure 5: Comparison of three types of transformer layers designed in this paper. [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Visual results for classical image ×4 SR on Urban100 dataset. 5.2 Evaluation of Hi-IR on Various IR tasks Image SR. For the classical image SR, we compared our Hi-IR with state-of-the-art SR models. The quantitative results are shown in Tab. 8. Aside from the 2nd-best …
Figure 7
Figure 7. Figure 7: Training one model for multiple degradation levels. [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 8
Figure 8. Figure 8: The columnar Hi-IR architecture [PITH_FULL_IMAGE:figures/full_fig_p020_8.png]
Figure 9
Figure 9. Figure 9: When the SR model is scale-up from Hi-IR-L to Hi-IR-B, the model Hi-IR-L converges slower than [PITH_FULL_IMAGE:figures/full_fig_p022_9.png]
Figure 10
Figure 10. Figure 10: Visual results for classical image ×4 SR on B100 dataset. F More Visual Results To further support the effectiveness and generalizability of the proposed Hi-IR intuitively. We provide more visual comparison in terms of image SR ( [PITH_FULL_IMAGE:figures/full_fig_p02…
Figure 11
Figure 11. Figure 11: Visual results for classical image ×4 SR on Manga109 dataset. 26 [PITH_FULL_IMAGE:figures/full_fig_p026_11.png]
Figure 12
Figure 12. Figure 12: Visual results for classical color image denoising on Urban100 dataset. The noise level is [PITH_FULL_IMAGE:figures/full_fig_p027_12.png]
Figure 13
Figure 13. Figure 13: Visual results for color image JPEG compression artifact removal on BSD500 dataset. The quality [PITH_FULL_IMAGE:figures/full_fig_p028_13.png]
Figure 14
Figure 14. Figure 14: Visual results for restoring images in adverse weather conditions. [PITH_FULL_IMAGE:figures/full_fig_p029_14.png]
Figure 15
Figure 15. Figure 15: Visual results for single image motion deblurring. The proposed method Hi-IR could recover [PITH_FULL_IMAGE:figures/full_fig_p030_15.png]
Figure 16
Figure 16. Figure 16: Visual results for single image motion deblurring. The proposed method Hi-IR could recover [PITH_FULL_IMAGE:figures/full_fig_p031_16.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

101 extracted references · 63 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

  2. [2]

    Defocus deblurring using dual-pixel data

    Abdullah Abuolaim and Michael S Brown. Defocus deblurring using dual-pixel data. In ECCV, pp.\ 111--126. Springer, 2020

    2020

  3. [3]

    Brown, and Peyman Milanfar

    Abdullah Abuolaim, Mauricio Delbracio, Damien Kelly, Michael S. Brown, and Peyman Milanfar. Learning to reduce defocus blur by realistically modeling dual-pixel data. In ICCV, 2021

    2021

  4. [4]

    NTIRE 2017 challenge on single image super-resolution: Dataset and study

    Eirikur Agustsson and Radu Timofte. NTIRE 2017 challenge on single image super-resolution: Dataset and study. In CVPRW, pp.\ 126--135, 2017

    2017

  5. [5]

    Densely residual laplacian super-resolution

    Saeed Anwar and Nick Barnes. Densely residual laplacian super-resolution. IEEE TPAMI, 44 0 (3): 0 1192--1204, 2020

    2020

  6. [6]

    Contour detection and hierarchical image segmentation

    Pablo Arbelaez, Michael Maire, Charless Fowlkes, and Jitendra Malik. Contour detection and hierarchical image segmentation. IEEE TPAMI, 33 0 (5): 0 898--916, 2010

    2010

  7. [7]

    Low-complexity single-image super-resolution based on nonnegative neighbor embedding

    Marco Bevilacqua, Aline Roumy, Christine Guillemot, and Marie Line Alberi-Morel. Low-complexity single-image super-resolution based on nonnegative neighbor embedding. In BMVC, 2012

    2012

  8. [8]

    Language models are few-shot learners

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learners. NeurIPS, 33: 0 1877--1901, 2020

    1901

Show all 101 references
  1. [9]

    Pre-trained image processing transformer

    Hanting Chen, Yunhe Wang, Tianyu Guo, Chang Xu, Yiping Deng, Zhenhua Liu, Siwei Ma, Chunjing Xu, Chao Xu, and Wen Gao. Pre-trained image processing transformer. In CVPR, pp.\ 12299--12310, 2021

    2021

  2. [10]

    Simple baselines for image restoration

    Liangyu Chen, Xiaojie Chu, Xiangyu Zhang, and Jian Sun. Simple baselines for image restoration. In ECCV, pp.\ 17--33. Springer, 2022 a

    2022

  3. [11]

    Activating more pixels in image super-resolution transformer

    Xiangyu Chen, Xintao Wang, Jiantao Zhou, Yu Qiao, and Chao Dong. Activating more pixels in image super-resolution transformer. In CVPR, pp.\ 22367--22377, 2023

    2023

  4. [12]

    Cross aggregation transformer for image restoration

    Zheng Chen, Yulun Zhang, Jinjin Gu, Linghe Kong, Xin Yuan, et al. Cross aggregation transformer for image restoration. NeurIPS, 35: 0 25478--25490, 2022 b

    2022

  5. [13]

    Rethinking coarse-to-fine approach in single image deblurring

    Sung-Jin Cho, Seo-Won Ji, Jun-Pyo Hong, Seung-Won Jung, and Sung-Jea Ko. Rethinking coarse-to-fine approach in single image deblurring. In ICCV, 2021

    2021

  6. [14]

    Conditional positional encodings for vision transformers

    Xiangxiang Chu, Zhi Tian, Bo Zhang, Xinlong Wang, and Chunhua Shen. Conditional positional encodings for vision transformers. In ICLR, 2022

    2022

  7. [15]

    Second-order attention network for single image super-resolution

    Tao Dai, Jianrui Cai, Yongbing Zhang, Shu-Tao Xia, and Lei Zhang. Second-order attention network for single image super-resolution. In CVPR, pp.\ 11065--11074, 2019

    2019

  8. [16]

    Image N et: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Image N et: A large-scale hierarchical image database. In CVPR, pp.\ 248--255. IEEE, 2009

    2009

  9. [17]

    Learning a deep convolutional network for image super-resolution

    Chao Dong, Chen Change Loy, Kaiming He, and Xiaoou Tang. Learning a deep convolutional network for image super-resolution. In ECCV, pp.\ 184--199. Springer, 2014

    2014

  10. [18]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, et al. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv...

    2010 arXiv

  11. [19]

    Quantization guided JPEG artifact correction

    Max Ehrlich, Larry Davis, Ser-Nam Lim, and Abhinav Shrivastava. Quantization guided JPEG artifact correction. In ECCV, pp.\ 293--309. Springer, 2020

    2020

  12. [20]

    Pointwise shape-adaptive dct for high-quality denoising and deblocking of grayscale and color images

    Alessandro Foi, Vladimir Katkovnik, and Karen Egiazarian. Pointwise shape-adaptive dct for high-quality denoising and deblocking of grayscale and color images. IEEE TIP, 16 0 (5): 0 1395--1411, 2007

    2007

  13. [21]

    Kodak lossless true color image suite

    Rich Franzen. Kodak lossless true color image suite. source: http://r0k. us/graphics/kodak, 4 0 (2), 1999

    1999

  14. [22]

    Deep joint demosaicking and denoising

    Micha \"e l Gharbi, Gaurav Chaurasia, Sylvain Paris, and Fr \'e do Durand. Deep joint demosaicking and denoising. ACM TOG, 35 0 (6): 0 1--12, 2016

    2016

  15. [23]

    Accurate, large minibatch sg d: training imagenet in 1 hour

    P Goyal. Accurate, large minibatch sg d: training imagenet in 1 hour. arXiv preprint arXiv:1706.02677, 2017

    2017 arXiv

  16. [24]

    Mamba: Linear-time sequence modeling with selective state spaces

    Albert Gu and Tri Dao. Mamba: Linear-time sequence modeling with selective state spaces. arXiv preprint arXiv:2312.00752, 2023

    2023 arXiv

  17. [25]

    MambaIR : A simple baseline for image restoration with state-space model

    Hang Guo, Jinmin Li, Tao Dai, Zhihao Ouyang, Xudong Ren, and Shu-Tao Xia. MambaIR : A simple baseline for image restoration with state-space model. arXiv preprint arXiv:2402.15648, 2024

    2024 arXiv

  18. [26]

    Residual learning for effective joint demosaicing-denoising

    Yu Guo, Qiyu Jin, Gabriele Facciolo, Tieyong Zeng, and Jean-Michel Morel. Residual learning for effective joint demosaicing-denoising. arXiv preprint arXiv:2009.06205, 2020

    2009 arXiv

  19. [27]

    Single image super-resolution from transformed self-exemplars

    Jia-Bin Huang, Abhishek Singh, and Narendra Ahuja. Single image super-resolution from transformed self-exemplars. In CVPR, pp.\ 5197--5206, 2015

    2015

  20. [28]

    Shuffle transformer: Rethinking spatial shuffle for vision transformer

    Zilong Huang, Youcheng Ben, Guozhong Luo, Pei Cheng, Gang Yu, and Bin Fu. Shuffle transformer: Rethinking spatial shuffle for vision transformer. arXiv preprint arXiv:2106.03650, 2021

    2021 arXiv

  21. [29]

    Towards flexible blind JPEG artifacts removal

    Jiaxi Jiang, Kai Zhang, and Radu Timofte. Towards flexible blind JPEG artifacts removal. In ICCV, pp.\ 4997--5006, 2021

    2021

  22. [30]

    Perceptual losses for real-time style transfer and super-resolution

    Justin Johnson, Alexandre Alahi, and Li Fei-Fei. Perceptual losses for real-time style transfer and super-resolution. In ECCV, pp.\ 694--711. Springer, 2016

    2016

  23. [31]

    Why warmup the learning rate? underlying mechanisms and improvements

    Dayal Singh Kalra and Maissam Barkeshli. Why warmup the learning rate? underlying mechanisms and improvements. arXiv preprint arXiv:2406.09405, 2024

    2024 arXiv

  24. [32]

    Scaling up GAN s for text-to-image synthesis

    Minguk Kang, Jun-Yan Zhu, Richard Zhang, Jaesik Park, Eli Shechtman, Sylvain Paris, and Taesung Park. Scaling up GAN s for text-to-image synthesis. In CVPR, pp.\ 10124--10134, 2023

    2023

  25. [33]

    Edge-based defocus blur estimation with adaptive scale selection

    Ali Karaali and Claudio Rosito Jung. Edge-based defocus blur estimation with adaptive scale selection. TIP, 2017

    2017

  26. [34]

    Beyond color difference: Residual interpolation for color image demosaicking

    Daisuke Kiku, Yusuke Monno, Masayuki Tanaka, and Masatoshi Okutomi. Beyond color difference: Residual interpolation for color image demosaicking. IEEE TIP, 25 0 (3): 0 1288--1300, 2016

    2016

  27. [35]

    Accurate image super-resolution using very deep convolutional networks

    Jiwon Kim, Jung Kwon Lee, and Kyoung Mu Lee. Accurate image super-resolution using very deep convolutional networks. In CVPR, pp.\ 1646--1654, 2016

    2016

  28. [36]

    MSSNet : Multi-scale-stage network for single image deblurring

    Kiyeon Kim, Seungyong Lee, and Sunghyun Cho. MSSNet : Multi-scale-stage network for single image deblurring. In ECCVW, pp.\ 524--539. Springer, 2022

    2022

  29. [37]

    Adam: A method for stochastic optimization

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980, 2014

    2014 arXiv

  30. [38]

    DeblurGAN-v2 : Deblurring (orders-of-magnitude) faster and better

    Orest Kupyn, Tetiana Martyniuk, Junru Wu, and Zhangyang Wang. DeblurGAN-v2 : Deblurring (orders-of-magnitude) faster and better. In ICCV, 2019

    2019

  31. [39]

    Deep defocus map estimation using domain adaptation

    Junyong Lee, Sungkil Lee, Sunghyun Cho, and Seungyong Lee. Deep defocus map estimation using domain adaptation. In CVPR, 2019

    2019

  32. [40]

    Iterative filter adaptive network for single image defocus deblurring

    Junyong Lee, Hyeongseok Son, Jaesung Rim, Sunghyun Cho, and Seungyong Lee. Iterative filter adaptive network for single image defocus deblurring. In CVPR, 2021

    2021

  33. [41]

    Heavy rain image restoration: Integrating physics model and conditional adversarial learning

    Ruoteng Li, Loong-Fah Cheong, and Robby T Tan. Heavy rain image restoration: Integrating physics model and conditional adversarial learning. In CVPR, pp.\ 1633--1642, 2019 a

    2019

  34. [42]

    All in one bad weather removal using architectural search

    Ruoteng Li, Robby T Tan, and Loong-Fah Cheong. All in one bad weather removal using architectural search. In CVPR, pp.\ 3175--3185, 2020

    2020

  35. [43]

    On efficient transformer and image pre-training for low-level vision

    Wenbo Li, Xin Lu, Jiangbo Lu, Xiangyu Zhang, and Jiaya Jia. On efficient transformer and image pre-training for low-level vision. arXiv preprint arXiv:2112.10175, 2021

    2021 arXiv

  36. [44]

    Efficient and explicit modelling of image hierarchies for image restoration

    Yawei Li, Yuchen Fan, Xiaoyu Xiang, Denis Demandolx, Rakesh Ranjan, Radu Timofte, and Luc Van Gool. Efficient and explicit modelling of image hierarchies for image restoration. In CVPR, pp.\ 18278--18289, 2023 a

    2023

  37. [45]

    LSDIR : A large scale dataset for image restoration

    Yawei Li, Kai Zhang, Jingyun Liang, Jiezhang Cao, Ce Liu, Rui Gong, Yulun Zhang, Hao Tang, Yun Liu, Denis Demandolx, et al. LSDIR : A large scale dataset for image restoration. In CVPRW, pp.\ 1775--1787, 2023 b

    2023

  38. [46]

    Feedback network for image super-resolution

    Zhen Li, Jinglei Yang, Zheng Liu, Xiaomin Yang, Gwanggil Jeon, and Wei Wu. Feedback network for image super-resolution. In CVPR, pp.\ 3867--3876, 2019 b

    2019

  39. [47]

    Blueprint separable residual network for efficient image super-resolution

    Zheyuan Li, Yingqi Liu, Xiangyu Chen, Haoming Cai, Jinjin Gu, Yu Qiao, and Chao Dong. Blueprint separable residual network for efficient image super-resolution. In CVPR, pp.\ 833--843, 2022

    2022

  40. [48]

    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. In ICCVW, pp.\ 1833--1844, 2021

    2021

  41. [49]

    Enhanced deep residual networks for single image super-resolution

    Bee Lim, Sanghyun Son, Heewon Kim, Seungjun Nah, and Kyoung Mu Lee. Enhanced deep residual networks for single image super-resolution. In CVPRW, pp.\ 1132--1140, 2017

    2017

  42. [50]

    DesnowNet : Context-aware deep network for snow removal

    Yun-Fu Liu, Da-Wei Jaw, Shih-Chia Huang, and Jenq-Neng Hwang. DesnowNet : Context-aware deep network for snow removal. IEEE TIP, 27 0 (6): 0 3064--3073, 2018

    2018

  43. [51]

    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 ICCV, pp.\ 10012--10022, 2021

    2021

  44. [52]

    Swin transformer v2: Scaling up capacity and resolution

    Ze Liu, Han Hu, Yutong Lin, Zhuliang Yao, Zhenda Xie, Yixuan Wei, Jia Ning, Yue Cao, Zheng Zhang, Li Dong, et al. Swin transformer v2: Scaling up capacity and resolution. In CVPR, pp.\ 12009--12019, 2022

    2022

  45. [53]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. In ICLR, 2018

    2018

  46. [54]

    Waterloo exploration database: New challenges for image quality assessment models

    Kede Ma, Zhengfang Duanmu, Qingbo Wu, Zhou Wang, Hongwei Yong, Hongliang Li, and Lei Zhang. Waterloo exploration database: New challenges for image quality assessment models. IEEE TIP, 26 0 (2): 0 1004--1016, 2016

    2016

  47. [55]

    Intriguing findings of frequency selection for image deblurring

    Xintian Mao, Yiming Liu, Fengze Liu, Qingli Li, Wei Shen, and Yan Wang. Intriguing findings of frequency selection for image deblurring. In AAAI, pp.\ 1905--1913, 2023

    1905

  48. [56]

    A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics

    David Martin, Charless Fowlkes, Doron Tal, and Jitendra Malik. A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics. In ICCV, volume 2, pp.\ 416--423. IEEE, 2001

    2001

  49. [57]

    Sketch-based manga retrieval using manga109 dataset

    Yusuke Matsui, Kota Ito, Yuji Aramaki, Azuma Fujimoto, Toru Ogawa, Toshihiko Yamasaki, and Kiyoharu Aizawa. Sketch-based manga retrieval using manga109 dataset. Multimedia Tools and Applications, 76 0 (20): 0 21811--21838, 2017

    2017

  50. [58]

    Image super-resolution with non-local sparse attention

    Yiqun Mei, Yuchen Fan, and Yuqian Zhou. Image super-resolution with non-local sparse attention. In CVPR, pp.\ 3517--3526, 2021

    2021

  51. [59]

    Deep multi-scale convolutional neural network for dynamic scene deblurring

    Seungjun Nah, Tae Hyun Kim, and Kyoung Mu Lee. Deep multi-scale convolutional neural network for dynamic scene deblurring. In CVPR, pp.\ 3883--3891, 2017

    2017

  52. [60]

    Single image super-resolution via a holistic attention network

    Ben Niu, Weilei Wen, Wenqi Ren, Xiangde Zhang, Lianping Yang, Shuzhen Wang, Kaihao Zhang, Xiaochun Cao, and Haifeng Shen. Single image super-resolution via a holistic attention network. In ECCV, pp.\ 191--207, 2020

    2020

  53. [61]

    Spatially-adaptive image restoration using distortion-guided networks

    Kuldeep Purohit, Maitreya Suin, AN Rajagopalan, and Vishnu Naresh Boddeti. Spatially-adaptive image restoration using distortion-guided networks. In ICCV, 2021

    2021

  54. [62]

    Attentive generative adversarial network for raindrop removal from a single image

    Rui Qian, Robby T Tan, Wenhan Yang, Jiajun Su, and Jiaying Liu. Attentive generative adversarial network for raindrop removal from a single image. In CVPR, pp.\ 2482--2491, 2018

    2018

  55. [63]

    Sharing key semantics in transformer makes efficient image restoration

    Bin Ren, Yawei Li, Jingyun Liang, Rakesh Ranjan, Mengyuan Liu, Rita Cucchiara, Luc Van Gool, Ming-Hsuan Yang, and Nicu Sebe. Sharing key semantics in transformer makes efficient image restoration. In NeurIPS, 2024

    2024

  56. [64]

    Bayesian-based iterative method of image restoration

    William Hadley Richardson. Bayesian-based iterative method of image restoration. JoSA, 62 0 (1): 0 55--59, 1972

    1972

  57. [65]

    Real-world blur dataset for learning and benchmarking deblurring algorithms

    Jaesung Rim, Haeyun Lee, Jucheol Won, and Sunghyun Cho. Real-world blur dataset for learning and benchmarking deblurring algorithms. In ECCV, pp.\ 184--201. Springer, 2020

    2020

  58. [66]

    Photorealistic text-to-image diffusion models with deep language understanding

    Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily L Denton, Kamyar Ghasemipour, Raphael Gontijo Lopes, Burcu Karagol Ayan, Tim Salimans, et al. Photorealistic text-to-image diffusion models with deep language understanding. NeurIPS, 35: 0 36479--36494, 2022

    2022

  59. [67]

    Live image quality assessment database release 2

    HR Sheikh. Live image quality assessment database release 2. http://live. ece. utexas. edu/research/quality, 2005

    2005

  60. [68]

    Human-aware motion deblurring

    Ziyi Shen, Wenguan Wang, Xiankai Lu, Jianbing Shen, Haibin Ling, Tingfa Xu, and Ling Shao. Human-aware motion deblurring. In ICCV, pp.\ 5572--5581, 2019

    2019

  61. [69]

    Just noticeable defocus blur detection and estimation

    Jianping Shi, Li Xu, and Jiaya Jia. Just noticeable defocus blur detection and estimation. In CVPR, 2015

    2015

  62. [70]

    Very deep convolutional networks for large-scale image recognition

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

    2015

  63. [71]

    Single image defocus deblurring using kernel-sharing parallel atrous convolutions

    Hyeongseok Son, Junyong Lee, Sunghyun Cho, and Seungyong Lee. Single image defocus deblurring using kernel-sharing parallel atrous convolutions. In ICCV, 2021

    2021

  64. [72]

    Scale-recurrent network for deep image deblurring

    Xin Tao, Hongyun Gao, Xiaoyong Shen, Jue Wang, and Jiaya Jia. Scale-recurrent network for deep image deblurring. In CVPR, 2018

    2018

  65. [73]

    LLaMA : Open and efficient foundation language models

    Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timoth \'e e Lacroix, Baptiste Rozi \`e re, Naman Goyal, Eric Hambro, Faisal Azhar, et al. LLaMA : Open and efficient foundation language models. arXiv preprint arXiv:2302.13971, 2023

    2023 arXiv

  66. [74]

    Stripformer: Strip transformer for fast image deblurring

    Fu-Jen Tsai, Yan-Tsung Peng, Yen-Yu Lin, Chung-Chi Tsai, and Chia-Wen Lin. Stripformer: Strip transformer for fast image deblurring. In ECCV, pp.\ 146--162. Springer, 2022 a

    2022

  67. [75]

    BANet : A blur-aware attention network for dynamic scene deblurring

    Fu-Jen Tsai, Yan-Tsung Peng, Chung-Chi Tsai, Yen-Yu Lin, and Chia-Wen Lin. BANet : A blur-aware attention network for dynamic scene deblurring. IEEE TIP, 31: 0 6789--6799, 2022 b

    2022

  68. [76]

    MAXIM : Multi-axis MLP for image processing

    Zhengzhong Tu, Hossein Talebi, Han Zhang, Feng Yang, Peyman Milanfar, Alan Bovik, and Yinxiao Li. MAXIM : Multi-axis MLP for image processing. In CVPR, pp.\ 5769--5780, 2022

    2022

  69. [77]

    TransWeather : Transformer-based restoration of images degraded by adverse weather conditions

    Jeya Maria Jose Valanarasu, Rajeev Yasarla, and Vishal M Patel. TransWeather : Transformer-based restoration of images degraded by adverse weather conditions. In CVPR, pp.\ 2353--2363, 2022

    2022

  70. [78]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, ukasz Kaiser, and Illia Polosukhin. Attention is all you need. NeurIPS, 30, 2017

    2017

  71. [79]

    Linformer: Self-attention with linear complexity

    Sinong Wang, Belinda Z Li, Madian Khabsa, Han Fang, and Hao Ma. Linformer: Self-attention with linear complexity. arXiv preprint arXiv:2006.04768, 2020

    2006 arXiv

  72. [80]

    ESRGAN : Enhanced super-resolution generative adversarial networks

    Xintao Wang, Ke Yu, Shixiang Wu, Jinjin Gu, Yihao Liu, Chao Dong, Yu Qiao, and Chen Change Loy. ESRGAN : Enhanced super-resolution generative adversarial networks. In ECCVW, pp.\ 0--0, 2018

    2018

  73. [81]

    Uformer: A general U -shaped transformer for image restoration

    Zhendong Wang, Xiaodong Cun, Jianmin Bao, Wengang Zhou, Jianzhuang Liu, and Houqiang Li. Uformer: A general U -shaped transformer for image restoration. In CVPR, pp.\ 17683--17693, 2022

    2022

  74. [82]

    Demosaicing based on directional difference regression and efficient regression priors

    Jiqing Wu, Radu Timofte, and Luc Van Gool. Demosaicing based on directional difference regression and efficient regression priors. IEEE TIP, 25 0 (8): 0 3862--3874, 2016

    2016

  75. [83]

    Random shuffle transformer for image restoration

    Jie Xiao, Xueyang Fu, Man Zhou, Hongjian Liu, and Zheng-Jun Zha. Random shuffle transformer for image restoration. In ICML, pp.\ 38039--38058, 2023

    2023

  76. [84]

    Vitae: Vision transformer advanced by exploring intrinsic inductive bias

    Yufei Xu, Qiming Zhang, Jing Zhang, and Dacheng Tao. Vitae: Vision transformer advanced by exploring intrinsic inductive bias. NeurIPS, 34: 0 28522--28535, 2021

    2021

  77. [85]

    Scaling up to excellence: Practicing model scaling for photo-realistic image restoration in the wild

    Fanghua Yu, Jinjin Gu, Zheyuan Li, Jinfan Hu, Xiangtao Kong, Xintao Wang, Jingwen He, Yu Qiao, and Chao Dong. Scaling up to excellence: Practicing model scaling for photo-realistic image restoration in the wild. In CVPR, 2024

    2024

  78. [86]

    Multi-stage progressive image restoration

    Syed Waqas Zamir, Aditya Arora, Salman Khan, Munawar Hayat, Fahad Shahbaz Khan, Ming-Hsuan Yang, and Ling Shao. Multi-stage progressive image restoration. In CVPR, pp.\ 14821--14831, 2021

    2021

  79. [87]

    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, pp.\ 5728--5739, 2022

    2022

  80. [88]

    On single image scale-up using sparse-representations

    Roman Zeyde, Michael Elad, and Matan Protter. On single image scale-up using sparse-representations. In Proceedings of International Conference on Curves and Surfaces, pp.\ 711--730. Springer, 2010

    2010

  81. [89]

    Accurate image restoration with attention retractable transformer

    Jiale Zhang, Yulun Zhang, Jinjin Gu, Yongbing Zhang, Linghe Kong, and Xin Yuan. Accurate image restoration with attention retractable transformer. In ICLR, 2022

    2022

  82. [90]

    Xformer: Hybrid x-shaped transformer for image denoising

    Jiale Zhang, Yulun Zhang, Jinjin Gu, Jiahua Dong, Linghe Kong, and Xiaokang Yang. Xformer: Hybrid x-shaped transformer for image denoising. arXiv preprint arXiv:2303.06440, 2023

    2023 arXiv

  83. [91]

    Beyond a G aussian denoiser: residual learning of deep CNN for image denoising

    Kai Zhang, Wangmeng Zuo, Yunjin Chen, Deyu Meng, and Lei Zhang. Beyond a G aussian denoiser: residual learning of deep CNN for image denoising. IEEE TIP, 26 0 (7): 0 3142--3155, 2017 a

    2017

  84. [92]

    Learning deep cnn denoiser prior for image restoration

    Kai Zhang, Wangmeng Zuo, Shuhang Gu, and Lei Zhang. Learning deep cnn denoiser prior for image restoration. In CVPR, pp.\ 3929--3938, 2017 b

    2017

  85. [93]

    Ffdnet: Toward a fast and flexible solution for cnn-based image denoising

    Kai Zhang, Wangmeng Zuo, and Lei Zhang. Ffdnet: Toward a fast and flexible solution for cnn-based image denoising. IEEE TIP, 27 0 (9): 0 4608--4622, 2018 a

    2018

  86. [94]

    Plug-and-play image restoration with deep denoiser prior

    Kai Zhang, Yawei Li, Wangmeng Zuo, Lei Zhang, Luc Van Gool, and Radu Timofte. Plug-and-play image restoration with deep denoiser prior. IEEE TPAMI, 2021

    2021

  87. [95]

    Color demosaicking by local directional interpolation and nonlocal adaptive thresholding

    Lei Zhang, Xiaolin Wu, Antoni Buades, and Xin Li. Color demosaicking by local directional interpolation and nonlocal adaptive thresholding. Journal of Electronic imaging, 20 0 (2): 0 023016, 2011

    2011

  88. [96]

    Image super-resolution using very deep residual channel attention networks

    Yulun Zhang, Kunpeng Li, Kai Li, Lichen Wang, Bineng Zhong, and Yun Fu. Image super-resolution using very deep residual channel attention networks. In ECCV, pp.\ 286--301, 2018 b

    2018

  89. [97]

    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 c

    2018

  90. [98]

    Residual non-local attention networks for image restoration

    Yulun Zhang, Kunpeng Li, Kai Li, Bineng Zhong, and Yun Fu. Residual non-local attention networks for image restoration. arXiv preprint arXiv:1903.10082, 2019

    1903 arXiv

  91. [99]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should ...

  92. [100]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@firs...

  93. [101]

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibset...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.