TSFormer: A Robust Framework for Efficient UHD Image Restoration

Chen Wu; Xin Su; Zhuoran Zheng

REVIEW 5 major objections 6 minor 1 cited by

TSFormer: A Robust Framework for Efficient UHD Image Restoration

T0 review · 5 major / 6 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 TSFormer claims to restore 4K images at 40 fps with just 3.38M parameters while beating far larger models on multiple UHD restoration benchmarks.

desk verdict The RMT 'trusted filtering' is never defined—Eq. 14 is a learned scalar with no spectral density—so the paper's central novelty is opaque, and the empirical gains are incremental. read the letter →

arxiv 2411.10951 v2 pith:DS7CQ4VV submitted 2024-11-17 cs.CV

Xin Su , Chen Wu , Zhuoran Zheng This is my paper · ORCID

classification cs.CV

keywords UHDimagerestorationtokensparsificationMin-psamplingrandommatrixtheorylightweighttransformerreal-time4Klow-lightenhancementdehazing

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 introduces TSFormer, a transformer for restoring ultra-high-definition images that deliberately lets very few tokens move through the network. Its central claim is that a dynamic token filter, Min-p sampling scaled by a trust scalar derived from the spectral density of the attention map, keeps only high-confidence features, so a 3.38M-parameter model can restore a 4K image in real time at about 40 fps while still matching or beating much larger restoration models. The paper reports state-of-the-art PSNR/SSIM on five UHD restoration tasks and shows that plugging the same filter into other transformer-based restorers cuts their FLOPs and runtime with minimal quality loss. If these claims hold, high-resolution restoration no longer requires heavy models, which matters for real-time applications on a single GPU.

What carries the argument

Min-p Sparse Attention (MSA), a sparse attention mechanism in which the attention map $M$ is computed by multiplying FFT-transformed query and key patches, then sparsified by retaining entries above $\text{Threshold} = p_{\text{base}}\max(M)$. Before thresholding, $M$ is treated as a random matrix, downsampled, and its spectral density is reduced to a scalar $\hat{\rho}$ by mean pooling and a sigmoid, rescaling the threshold via $\hat{\text{Threshold}} = \hat{\rho} \cdot \text{Threshold}$. The scalar is meant to raise the threshold when the attention matrix is noisy or random, so that unstable tokens are filtered. MSA sits inside Trusted Sparse Blocks together with a feed-forward network, and the whole U-shaped network uses Feature Fusion Blocks to merge encoder-decoder features.

What would settle it

Fix all hyperparameters and replace the learned trust scalar $\hat{\rho}$ with a constant, or inject controlled white noise into the attention map and check whether $\hat{\rho}$ rises with noise variance; if the restored-image PSNR does not drop when $\hat{\rho}$ is frozen, or if the scalar does not track injected noise, the claimed random-matrix-theory gate is not the cause of the robustness gains.

Watch

Extended reading notes

Core claim

The central discovery, as the authors present it, is that most tokens in an attention map for UHD restoration are not worth attending to, and that a probability-based threshold with a random-matrix-theory trust adjustment selects the rest. Concretely, attention scores are computed via FFT, then any score below $\text{Threshold} = p_{\text{base}} \cdot \max(M)$ is zeroed, and the threshold is multiplied by $\hat{\rho}$, a scalar obtained by downsampling the attention matrix, computing its spectral density, applying mean pooling and a sigmoid. The result, Min-p Sparse Attention, repeats inside every block of a U-shaped encoder-decoder. Across UHD low-light, dehazing, deblurring, deraining, and desnowing benchmarks, the model reports the best PSNR and SSIM among the compared methods, with 3.38M parameters and a reported 4K inference speed near 40 fps, and the same MSA unit accelerates LLFormer, UHDformer, and UHDFour when added to them.

Load-bearing premise

The claim rests on the assumption that the scalar computed from the spectral density of the downsampled attention matrix, after mean pooling and a sigmoid, genuinely measures token uncertainty and noise, so that multiplying the Min-p threshold by it suppresses noise and improves robustness; if that scalar is not a real noise measure, the trusted-filtering results are unsupported.

Editorial extensions

If this is right

A 4K image can be restored in about 25 ms on a single GPU, which makes real-time UHD video restoration feasible.
Other transformer-based restoration models can adopt MSA and cut FLOPs and runtime by roughly 20-25 percent while approximately maintaining PSNR and SSIM, as demonstrated on LLFormer, UHDformer, and UHDFour.
Restoration quality does not need to scale with parameter count on these benchmarks: 3.38M parameters outperforms models with many times more parameters.
The same architecture handles multiple degradations, including low light, haze, blur, rain, and snow, without task-specific heads, suggesting a single all-in-one model can serve UHD restoration.

Reading between the lines

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

One immediate independent test is to replace the learned scalar $\hat{\rho}$ with a constant while keeping everything else fixed; if the PSNR gain on UHD-Haze largely disappears, the Min-p threshold alone may be doing most of the work.
The sparse-attention gate could transfer to other high-resolution dense prediction tasks, such as UHD video super-resolution or compression artifact removal, where throughput is the main bottleneck.
A stricter version of the trusted mechanism would derive the threshold from a formal confidence bound on the eigenvalue distribution rather than a sigmoid of a mean-pooled spectral density; such a version would make the random-matrix-theory connection testable in isolation.

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.

Desk Editor's Note

read the letter

Two things you should know. First, the paper's stated novelty—random-matrix-theory trusted filtering—is not actually in the paper. Eq. (14) defines the trusted threshold as rho times the Min-p threshold, where rho comes from a mean-pooling layer and sigmoid applied to the 'spectral density' of the attention matrix. No spectral density is ever computed or defined; no eigenvalue analysis appears; no theorem is invoked. The stress-test note is right: this is an opaque placeholder, not a mechanism. Second, the actual system is a reasonably clean combination of Min-p sampling (borrowed from LLM decoding) with FFT-based attention, on a lightweight U-shaped net. That part is coherent, and the plug-in experiments in Table 10, if the numbers could be trusted, would be a useful engineering data point.

What the paper does well: it runs five UHD restoration tasks, compares against a broad set of baselines, and the MSA idea is simple enough to be dropped into other transformers. The writing is readable. The architecture figure is clear. The Min-p attention sparsification is a legitimate idea, and the ablation in Table 6 shows it helps over top-k by a couple tenths of a dB.

Now the soft spots, in proportion. The central flaw is load-bearing: if the RMT claim is removed, the paper is an incremental engineering entry—gains over UHDDIP are under 0.3 dB, often under 0.1, and TSFormer uses about four times the parameters. The 4K 40fps claim is extrapolated from a 1024x1024 timing on a 2080Ti; naive scaling gives roughly 10–20fps at 4K, and no 4K timing is reported. The tables have internal inconsistencies: Table 10 gives UHDFormer 34.55M parameters while Table 1 and Table 4 list UHDformer at 0.34M; the names UHD, UHDFour, and UHDformer are shuffled. The FED and ISA ablations in Table 9 get two sentences each and implausible runtimes. No code, no data, and a handful of references look unreliable. These are not minor nits; they make the empirical claims hard to verify.

Who is this for? A reader focused on efficient UHD restoration might pick up the Min-p + sparse attention idea from this paper and try it in their own model, but they would want code and honest numbers before trusting it.

Recommendation: If I were an editor, I would not desk reject outright—there is a real architectural idea here—but I would send it to review only with the expectation that the RMT claim be removed or actually implemented, code be released, and the timing claim be remade on real 4K data. In its current form, I would not accept or recommend anyone build on it.

Referee Report

5 major / 6 minor

Summary. The manuscript proposes TSFormer, a lightweight transformer for UHD image restoration that combines Min-p sampling with a so-called trusted learning mechanism based on random matrix theory (RMT) to sparsify attention. The architecture is an encoder-decoder with Trusted Sparse Blocks containing Min-p Sparse Attention (MSA), frequency-domain attention computation, and feature fusion blocks. Experiments are reported on five UHD tasks (low-light enhancement, dehazing, deblurring, deraining, desnowing) with claims of state-of-the-art PSNR/SSIM and real-time 4K inference (40 fps) at 3.38M parameters. The paper also claims that MSA can be plugged into other transformer-based restoration models to improve efficiency.

Significance. If the claims were substantiated, the combination of Min-p sparsification and an RMT-based trusted filter could offer a useful efficiency/quality trade-off for UHD restoration, and the plug-and-play MSA study would broaden its impact. The paper contains a wide set of experiments across five UHD benchmarks and a generalization study on three existing models, which is a strength. However, the central RMT mechanism is not actually defined (Eq. 14 leaves the spectral density unspecified), and the experimental tables contain internal inconsistencies. As a result, the significance of the work cannot be assessed from the current text: the core novelty is a placeholder, and the empirical validation is unreliable in places.

major comments (5)

[Section 3.3, Eq. (14)] The trusted learning mechanism is not defined. Equation (14) sets \hat{Threshold} = \hat{\rho} \cdot Threshold, where \hat{\rho} is said to be generated from 'the spectral density \rho of the random matrix M' through 'a mean pooling layer and a sigmoid function'. No formula for \rho is given, no matrix ensemble is specified, the pooling/sigmoid composition is not described, and no learning rule or loss term is provided for \hat{\rho}. The claim that 'If M is more random, it may contain more noise and therefore requires a higher threshold' is an assertion without derivation. Because this mechanism is the paper's claimed novelty for robustness and generalization, the central contribution is not a well-defined scientific claim, and the experiments cannot be interpreted as testing it.
[Section 6 and Table 7] The real-time 4K claim of 40 fps is not supported by the reported measurements. Table 7 gives an inference time of 0.012 s at 1024×1024 resolution, which corresponds to about 83 fps at that resolution. A 4K frame (3840×2160) has roughly 8.3 times as many pixels, so a linear extrapolation gives about 10 fps, not 40 fps. The paper's statement 'Without considering I/O operations, we infer that a 4K image requires 40 fps' needs a direct measurement on 4K inputs or a careful scaling analysis accounting for patch-based processing and resolution-dependent components.
[Table 10] The models in Table 10 appear to be mislabeled, undermining the plug-and-play claims in Section 7. 'UHDFormer' is listed with 34.55M parameters and 36.61 dB PSNR, but Table 1 reports UHDformer with 0.34M parameters, and Tables 5 and 8 report UHDformer achieving 36.61 dB on UHD-Rain and UHD-Snow. 'UHDFour' is listed with 41.56 dB PSNR, a value that matches UHDDIP in Tables 5 and 8, not UHDFour's 26.23 dB on UHD-LL in Table 1. These inconsistencies make it impossible to verify the claimed efficiency gains and minimal performance loss when integrating MSA.
[Tables 2, 4, 5, 8] The reported performance gains over the strongest baseline UHDDIP are small: 0.01 dB on UHD-Blur, 0.19 dB on UHD-Haze, 0.22 dB on UHD-Rain, and 0.26 dB on UHD-Snow. No error bars, confidence intervals, or repeated-run statistics are provided. Differences of this magnitude are within typical training variation for image restoration models, so the claim of state-of-the-art robustness and generalization is not statistically substantiated.
[Section 5.2 and Table 9] The ablation against FED and ISA is not reproducible. FED and ISA are each described in two sentences, with no specification of patch size, how the eigenvalue threshold \tau is chosen, how the strategies are integrated into the MSA module, or how the runtime is measured. The runtime difference between FED (0.45 s) and the proposed method (0.012 s) is over 37×, but no complexity analysis is given to explain where this speed-up comes from. Without implementation details, the comparison cannot be verified or reproduced.

minor comments (6)

[Section 3.2] The sentence 'A decoder is a mirrored operation of a decoder' should probably read 'a mirrored operation of the encoder'.
[Section 3.3, Eqs. (10)-(11)] The shapes of QFFT, KFFT, and M are not specified, so it is unclear whether the FFT is applied along the spatial dimension or the token dimension, and what the resulting attention matrix size is.
[Section 3.3, Eq. (13)] The operation in Eq. (13) is a deterministic hard threshold, not the stochastic 'Min-p sampling' used in LLM decoding; the terminology should be clarified to avoid confusion.
[Figure 7] The 'cumulative probability distribution plots' are not explained; the axes are not labeled, and the relation of the plots to the sampling methods is unclear.
[References] Several references are inconsistent or appear to use placeholder-like arXiv numbers: [15] is 'arXiv:2303.12345', [22] is 'arXiv:2301.12345', and two different entries ([5] and [30], [21] and [22]) describe the same methods (SFNet and DehazeFormer) with different authors and titles. The reference list should be checked.
[Section 7 heading] The heading 'MSA of Potential' is likely a typo and should be revised.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's contribution is an empirical network design and benchmark evaluation, not a derivation that reduces to its inputs.

full rationale

After walking the paper's derivation chain, no claim is equivalent to its input by construction. The only candidate is Eq. 14, where the trusted threshold is defined as ρ̂ · Threshold, with ρ̂ generated from an unspecified spectral density of the attention matrix M. The paper asserts a heuristic relation ('If M is more random, it may contain more noise and therefore requires a higher threshold') but never defines the spectral density computation nor proves any connection to random matrix theory. This is an underspecified, unfalsifiable design assumption, which is a validity and reproducibility concern, not circularity: no equation is defined in terms of the target result, no fitted parameter is relabeled as a prediction, and no self-citation or imported uniqueness theorem carries the argument. The benchmark improvements are ordinary trained-model results on the same distributions, and the MSA plug-and-play gains are measured empirically rather than derived from the trusted mechanism. Because the hard rule requires exhibiting a specific reduction and none can be exhibited, the circularity score is 0.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The central mechanism rests on one hand-set hyperparameter (p_base) and one learned per-example scaling (rho_hat), both tuned on the target data. The random-matrix-theory citation is not operationalized; no eigenvalue statistics are used in the final model.

free parameters (2)

p_base (Min-p threshold ratio) = not reported
Hyperparameter in Eq. (1) and (12) that controls attention sparsity; selected by hand and never swept in ablations.
rho_hat threshold multiplier (mean-pooling + sigmoid weights) = learned end-to-end
The trusted scalar in Eq. (14) is produced by trainable network parameters; it is fitted on the same restoration datasets, so the robustness mechanism is not a parameter-free derivation.

assumptions (3)

ad hoc to paper Spectral density of the (downsampled) attention matrix, summarized by a sigmoid of its mean, reliably measures token uncertainty.
Invoked in Sec. 3.3 Eq. (14); no derivation or citation is given for this specific operation; it is a learned gate.
domain assumption Element-wise frequency-domain product of query and key patches followed by IFFT yields an attention map suitable for thresholding.
Eq. (10)-(11) in Sec. 3.3; standard FFT correlation, but no proof that it approximates attention scores for the sparsification purpose.
domain assumption Min-p thresholding based on the maximum logit retains the most informative tokens for restoration.
Borrowed from LLM token sampling (Sec. 3.1, Eq. 1-2); assumed to transfer to image attention maps.

how reviews work

0 comments

Cite this review

Pith. "Pith review of TSFormer: A Robust Framework for Efficient UHD Image Restoration." pith.science (2026). https://pith.science/paper/DS7CQ4VV

@misc{pith2026241110951,
  author       = {Pith},
  title        = {Pith review of: TSFormer: A Robust Framework for Efficient UHD Image Restoration},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DS7CQ4VV}},
  note         = {Machine review of arXiv:2411.10951}
}

abstract

Ultra-high-definition (UHD) image restoration is vital for applications demanding exceptional visual fidelity, yet existing methods often face a trade-off between restoration quality and efficiency, limiting their practical deployment. In this paper, we propose TSFormer, an all-in-one framework that integrates \textbf{T}rusted learning with \textbf{S}parsification to boost both generalization capability and computational efficiency in UHD image restoration. The key is that only a small amount of token movement is allowed within the model. To efficiently filter tokens, we use Min-$p$ with random matrix theory to quantify the uncertainty of tokens, thereby improving the robustness of the model. Our model can run a 4K image in real time (40fps) with 3.38 M parameters. Extensive experiments demonstrate that TSFormer achieves state-of-the-art restoration quality while enhancing generalization and reducing computational demands. In addition, our token filtering method can be applied to other image restoration models to effectively accelerate inference and maintain performance.

Figures

Figures reproduced from arXiv: 2411.10951 by the authors.

**Figure 1.** Illustration of the Min-p sampling and stability filtering process.(a) The original input image used as a reference. (b) Result after Min-p sampling, where high-probability regions are retained and highlighted in a distinct color, indicating sparsified yet significant areas. (c) Result after stability filtering, where only the stable, high-confidence features are preserved, with stable regions marked in red and un… view at source ↗

**Figure 2.** The overall architecture of the proposed TSFormer for UHD image restoration, which main consists of Trusted Sparse Blocks [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Comparison of Low-light Enhancement Methods. The top row is from the UHD-LL dataset, and the bottom is from the UHD [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Figures from the paper (7 more)

**Figure 4.** Figure 4: Image deblurring on UHD-Blur. TSFormer is able to generate deblurring results with sharper structures. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

**Figure 5.** Figure 5: Image dehazing on UHD-Haze. TSFormer is capable of producing clearer results. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

**Figure 7.** Figure 7: Cumulative probability distribution plots for different sampling methods. Min-p Sampling with trusted mechanism (c) demon [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]

**Figure 8.** Figure 8: Image deraining on UHD-Snow. Visual comparisons of different methods on the UHD-Snow dataset. TSFormer effectively [PITH_FULL_IMAGE:figures/full_fig_p006_8.png]

**Figure 9.** Figure 9: Comparison of different sampling methods for UHD image dehazing. (Left) The full-sized input image. (Right) Cropped samples [PITH_FULL_IMAGE:figures/full_fig_p007_9.png]

**Figure 10.** Figure 10: Object detection visual comparison on the DarkFace [ [PITH_FULL_IMAGE:figures/full_fig_p007_10.png]

**Figure 11.** Figure 11: Comparison of real-world low-light enhancement results with and without MSA. MSA can improve generalization and feature [PITH_FULL_IMAGE:figures/full_fig_p007_11.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Deep Learning-Driven Ultra-High-Definition Image Restoration: A Survey
cs.CV 2025-05 conditional novelty 3.0 of 10

A structured survey of UHD image restoration with new comparative experiments on backbones, sampling, losses, and state-of-the-art methods across six degradation types.

Reference graph

Works this paper leans on

42 extracted references · 34 canonical work pages · cited by 1 Pith paper

[1]

Estimating error probabilities of sparse representations

Guillaume Bellec, Johannes Brunner, Carsten Rother, and Damian Tschannen. Estimating error probabilities of sparse representations. IEEE Transactions on Information Theory, 69(3):1150–1165, 2023. 1, 2

work page 2023
[2]

Ran- dom matrix theory and machine learning: a big data perspec- tive

Jo ¨el Bun, Jean-Philippe Bouchaud, and Marc Potters. Ran- dom matrix theory and machine learning: a big data perspec- tive. Physics Reports, 666:1–109, 2017. 2

work page 2017
[3]

Ultra-high-definition video streaming and surveillance: Challenges and trends

Li Chen, Jin Xu, and Wei Zhang. Ultra-high-definition video streaming and surveillance: Challenges and trends. IEEE Communications Magazine, 59(7):18–24, 2021. 1

work page 2021
[4]

Random matrix theory for big data analytics in large-scale systems

Romain Couillet and M ´erouane Debbah. Random matrix theory for big data analytics in large-scale systems. Pro- ceedings of the IEEE, 106(8):1420–1450, 2018. 2

work page 2018
[5]

Sfnet: A shape and feature fusion network for high-resolution image restora- tion

Jiahao Cui, Wei Li, and Qiang Zhang. Sfnet: A shape and feature fusion network for high-resolution image restora- tion. IEEE Transactions on Image Processing , 32:1961– 1975, 2023. 8

work page 1961
[6]

Separable-patch integration network for uhd video deblurring

Y Deng, L Wang, and Q Zhang. Separable-patch integration network for uhd video deblurring. IEEE Transactions on Computational Imaging, 7(2):123–136, 2021. 2

work page 2021
[7]

Eigenvalues and condition numbers of ran- dom matrices

Alan Edelman. Eigenvalues and condition numbers of ran- dom matrices. SIAM Review, 30(3):507–536, 1988. 1, 2

work page 1988
[8]

Hierarchical neural story generation

Angela Fan, Mike Lewis, and Yann Dauphin. Hierarchical neural story generation. arXiv preprint arXiv:1805.04833 ,

arXiv

Show all 42 references

[9]

Anal- ysis of the empirical eigenvalue distribution of large dimen- sional information-plus-noise type matrices

Walid Hachem, Philippe Loubaton, and Jamal Najim. Anal- ysis of the empirical eigenvalue distribution of large dimen- sional information-plus-noise type matrices. IEEE Transac- tions on Information Theory, 53(3):1057–1079, 2007. 2

2007
[10]

Robustness in high- dimensional data processing with random matrix theory

Rui Huang, Lei Wang, and Qing Zhang. Robustness in high- dimensional data processing with random matrix theory. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition (CVPR) , pages 5213–5221,
[11]

Scope of va- lidity of psnr in image/video quality assessment

Quan Huynh-Thu and Mohammed Ghanbari. Scope of va- lidity of psnr in image/video quality assessment. Electronics letters, 44(13):800–801, 2008. 5

2008
[12]

Fftformer: A lightweight transformer for image restoration.arXiv preprint arXiv:2302.05645, 2023

Shuai Kong, Xiaoyuan Liu, and Yang Wang. Fftformer: A lightweight transformer for image restoration.arXiv preprint arXiv:2302.05645, 2023. 5, 8

2023 arXiv
[13]

Embedding fourier for ultra-high-definition low-light image enhance- ment

Chenyang Li, Chengze Guo, Ming Zhou, Zihan Liang, Shuang Zhou, Rui Feng, and Chen Change Loy. Embedding fourier for ultra-high-definition low-light image enhance- ment. In International Conference on Learning Represen- tations (ICLR), 2023. 5

2023
[14]

Embedding fourier for ultra-high-definition low-light image enhancement

Chongyi Li, Chun-Le Guo, Man Zhou, Zhexin Liang, Shangchen Zhou, Ruicheng Feng, and Chen Change Loy. Embedding fourier for ultra-high-definition low-light image enhancement. In ICLR, 2023. 5

2023
[15]

Uhd-four: Ultra-high- definition image restoration via fourier transform

Yi Li, Yan Chen, and Tao Huang. Uhd-four: Ultra-high- definition image restoration via fourier transform. arXiv preprint arXiv:2303.12345, 2023. 5, 8

2023 arXiv
[16]

Swinir: Image restoration us- ing swin transformer

Jingyun Liang, Jiezhang Cao, Guolei Sun, Kai Zhang, Luc Van Gool, and Radu Timofte. Swinir: Image restoration us- ing swin transformer. InProceedings of the IEEE/CVF Inter- national Conference on Computer Vision (ICCV) Workshops, pages 1833–1844, 2021. 5

2021
[17]

High- resolution medical image reconstruction using deep learning: Promises and challenges

Yun Liu, Jun Cheng, Yicheng Zhang, and Yan Wang. High- resolution medical image reconstruction using deep learning: Promises and challenges. IEEE Transactions on Medical Imaging, 39(5):1390–1400, 2020. 1

2020
[18]

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/CVF International Conference on Computer Vision (ICCV), pages 10012–10022, 2021. 1, 2

2021
[19]

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 Proceedings of the IEEE Conference on Com- puter Vision and Pattern Recognition (CVPR), pages 3883– 3891, 2017. 5

2017
[20]

Min p sampling: Balancing creativity and coherence at high temperature

Minh Nguyen, Andrew Baker, Andreas Kirsch, and Clement Neo. Min p sampling: Balancing creativity and coherence at high temperature. arXiv preprint arXiv:2407.01082, 2024. 2

2024
[21]

Dehazeformer: Transformer-based image dehazing

Lei Song, Yifan Feng, Xin Wang, Hui Zhang, and Bin Xu. Dehazeformer: Transformer-based image dehazing. IEEE Transactions on Circuits and Systems for Video Technology, 33(2):1234–1248, 2023. 8

2023
[22]

Dehazeformer: Transformer-based image dehazing via multi-scale feature aggregation

Wei Song, Kai Zhang, and Liang Wang. Dehazeformer: Transformer-based image dehazing via multi-scale feature aggregation. arXiv preprint arXiv:2301.12345, 2023. 5

2023 arXiv
[23]

Topics in Random Matrix Theory

Terence Tao. Topics in Random Matrix Theory . American Mathematical Society, 2012. 1, 2

2012
[24]

Strip- former: Strip transformer for fast image processing

Yi-Hsin Tsai, Ming-Hsuan Yang, and Yu-Chuan Lin. Strip- former: Strip transformer for fast image processing. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 850–859, 2022. 5, 8

2022
[25]

Correlation matching transformation transformers for uhd image restora- tion

Chenhui Wang, Jiamin Pan, Wei Wang, Gaoyang Fu, Shuai Liang, Meng Wang, Xiaoming Wu, and Jie Liu. Correlation matching transformation transformers for uhd image restora- tion. arXiv preprint arXiv:2406.00629, 2024. 5

2024 arXiv
[26]

Llformer: Transformer-based low-light image enhancement for uhd

J Wang, X Li, and H Zhang. Llformer: Transformer-based low-light image enhancement for uhd. IEEE Transactions on Neural Networks and Learning Systems, 34(5):987–1000,
[27]

Uhdformer: A lightweight model for ultra-high-definition image restoration

J Wang, L Zhao, and S Chen. Uhdformer: A lightweight model for ultra-high-definition image restoration. IEEE Transactions on Image Processing , 35:789–803, 2024. 5, 8

2024
[28]

Ultra-high-definition restoration: New benchmarks and a dual interaction prior- driven solution

Liyan Wang, Cong Wang, Jinshan Pan, Weixiang Zhou, Xi- aoran Sun, Wei Wang, and Zhixun Su. Ultra-high-definition restoration: New benchmarks and a dual interaction prior- driven solution. arXiv preprint arXiv:2406.13607, 2024. 5

2024 arXiv
[29]

Ultra-high-definition low-light image enhancement: A benchmark and transformer-based method

Tianyu Wang, Kai Zhang, Tao Shen, Wenhan Luo, Bernd Stenger, and Tong Lu. Ultra-high-definition low-light image enhancement: A benchmark and transformer-based method. In Proceedings of the AAAI Conference on Artificial Intelli- gence, pages 2302–2310, 2023. 5

2023
[30]

Sfnet: Spatial- frequency network for image restoration

Tianyu Wang, Kai Zhang, and Tong Lu. Sfnet: Spatial- frequency network for image restoration. arXiv preprint arXiv:2401.12345, 2024. 5

2024 arXiv
[31]

Ultra-high-definition image restoration with dual interaction priors

Xiaoyun Wang, Fei Zhou, and Li Chen. Ultra-high-definition image restoration with dual interaction priors. In Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024. 1, 2, 5

2024
[32]

Image quality assessment: from error visibility to structural similarity

Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Si- moncelli. Image quality assessment: from error visibility to structural similarity. IEEE transactions on image processing, 13(4):600–612, 2004. 5

2004
[33]

Uformer: A general u- shaped transformer for image restoration

Zhendong Wang, Xiaodong Cun, Jianmin Bao, Wenqiang Zhou, Jing Liu, and Hanzi Li. Uformer: A general u- shaped transformer for image restoration. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 17683–17693, 2022. 5, 8

2022
[34]

One-shot ultra-high-resolution image synthesis with our-gan

Lining Xu, Xianfang Wang, and Zhiyong Liu. One-shot ultra-high-resolution image synthesis with our-gan. arXiv preprint arXiv:2202.13799, 2022. 1, 2

2022 arXiv
[35]

Uncertainty and robust- ness in deep learning models

Hao Yang, Feng Li, and Ming Chen. Uncertainty and robust- ness in deep learning models. IEEE Transactions on Neural Networks and Learning Systems, 2024. To appear. 2

2024
[36]

Advanced techniques in high-resolution image processing

Min Yang, Xia Li, and Ling Zhao. Advanced techniques in high-resolution image processing. IEEE Transactions on Image Processing, 33(4):2255–2268, 2024. 1, 2

2024
[37]

Advancing image under- standing in poor visibility environments: A collective bench- mark study

Wenhan Yang, Ye Yuan, Wenqi Ren, Jiaying Liu, Wal- ter J Scheirer, Zhangyang Wang, Taiheng Zhang, Qiaoyong Zhong, Di Xie, Shiliang Pu, et al. Advancing image under- standing in poor visibility environments: A collective bench- mark study. IEEE Transactions on Image Processing ...

2020
[38]

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. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 572...

2022
[39]

Wave-mamba: A wavelet-based state space model for ultra-high-definition im- age restoration

Ming Zhao, Yu Chen, and Rui Xu. Wave-mamba: A wavelet-based state space model for ultra-high-definition im- age restoration. arXiv preprint arXiv:2408.01276, 2024. 1, 2

2024 arXiv
[40]

Multi-guided bilateral upsam- pling for uhd image dehazing

X Zheng, Y Li, and Z Wang. Multi-guided bilateral upsam- pling for uhd image dehazing. IEEE Transactions on Image Processing, 30:456–468, 2021. 2

2021
[41]

Uhd: A benchmark for ultra-high-definition image restoration

Yi Zheng, Xi Wang, and Shuai Li. Uhd: A benchmark for ultra-high-definition image restoration. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2021. 5, 8

2021
[42]

Self- sampling for probabilistic sparsification in transformers

Jian Zhou, Mingkui Tan, Hao Li, and Sen Zhao. Self- sampling for probabilistic sparsification in transformers. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 4882–4890, 2020. 2

2020

Pith tools

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

[1] [1]

Estimating error probabilities of sparse representations

Guillaume Bellec, Johannes Brunner, Carsten Rother, and Damian Tschannen. Estimating error probabilities of sparse representations. IEEE Transactions on Information Theory, 69(3):1150–1165, 2023. 1, 2

work page 2023

[2] [2]

Ran- dom matrix theory and machine learning: a big data perspec- tive

Jo ¨el Bun, Jean-Philippe Bouchaud, and Marc Potters. Ran- dom matrix theory and machine learning: a big data perspec- tive. Physics Reports, 666:1–109, 2017. 2

work page 2017

[3] [3]

Ultra-high-definition video streaming and surveillance: Challenges and trends

Li Chen, Jin Xu, and Wei Zhang. Ultra-high-definition video streaming and surveillance: Challenges and trends. IEEE Communications Magazine, 59(7):18–24, 2021. 1

work page 2021

[4] [4]

Random matrix theory for big data analytics in large-scale systems

Romain Couillet and M ´erouane Debbah. Random matrix theory for big data analytics in large-scale systems. Pro- ceedings of the IEEE, 106(8):1420–1450, 2018. 2

work page 2018

[5] [5]

Sfnet: A shape and feature fusion network for high-resolution image restora- tion

Jiahao Cui, Wei Li, and Qiang Zhang. Sfnet: A shape and feature fusion network for high-resolution image restora- tion. IEEE Transactions on Image Processing , 32:1961– 1975, 2023. 8

work page 1961

[6] [6]

Separable-patch integration network for uhd video deblurring

Y Deng, L Wang, and Q Zhang. Separable-patch integration network for uhd video deblurring. IEEE Transactions on Computational Imaging, 7(2):123–136, 2021. 2

work page 2021

[7] [7]

Eigenvalues and condition numbers of ran- dom matrices

Alan Edelman. Eigenvalues and condition numbers of ran- dom matrices. SIAM Review, 30(3):507–536, 1988. 1, 2

work page 1988

[8] [8]

Hierarchical neural story generation

Angela Fan, Mike Lewis, and Yann Dauphin. Hierarchical neural story generation. arXiv preprint arXiv:1805.04833 ,

arXiv