Pith. sign in

REVIEW 3 major objections 5 minor 62 references

FiDeSR: High-Fidelity and Detail-Preserving One-Step Diffusion Super-Resolution

T0 review · 3 major / 5 minor · reviewed 2026-08-02 · deepseek-v4-flash

Pith's one-line read One-step diffusion super-resolution can raise perceptual quality and structural fidelity together, when training is error-weighted and the latent is frequency-refined at inference.

desk verdict Useful one-step SR engineering, but the headline numbers are partly tuned on the test set — the central SOTA claim needs re-validation before I'd trust it. read the letter →

arxiv 2603.02692 v1 pith:L5K4MKKI submitted 2026-03-03 cs.CV

classification cs.CV
keywords imagesuper-resolutionone-stepdiffusionlatentmodelsdetail-awareweightingresidualrefinementfrequencyinjectionreal-worldperception-distortiontrade-off
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 sets out to establish that the perception–distortion trade-off that has forced diffusion-based super-resolution to run tens or hundreds of denoising steps can be broken with a single step: FiDeSR's restored images are reported to be both more faithful in structure and more realistic in texture than those of existing diffusion methods. It diagnoses why one-step diffusion SR fails — low-frequency inconsistency from latent compression and under-produced high-frequency detail from truncated denoising — and addresses each failure with a named mechanism. Detail-aware Weighting concentrates training loss on regions where prediction error is largest; the Latent Residual Refinement Block corrects the diffusion network's coarse residual prediction; and the Latent Frequency Injection Module reinjects low- and high-frequency content at inference with user-adjustable strength. On real-world benchmarks the paper reports the best perceptual scores among nine diffusion methods at a single diffusion step, suggesting that efficient one-step restoration need not surrender fidelity.

What carries the argument

The load-bearing object is the residual latent restoration identity zr = zL − r', which reduces super-resolution to predicting how much to subtract from the low-quality latent. FiDeSR upgrades this identity in two ways. LRRB replaces the single coarse residual with a refined one, r' = r + ∆r, where ∆r is produced by a stack of residual-in-residual dense blocks operating on the concatenation of the LQ latent and the initial residual — a learned correction that directly targets the instability of one-step noise prediction. LFIM then modifies the latent itself at inference: z ← z + α·M_sp·M_ch·∆_LP for low frequencies, and z ← z + β·M_sp·M_ch·∆_HP for high frequencies, where M_sp and M_ch are s

What would settle it

Re-evaluate FiDeSR on DRealSR and RealSR with LFIM parameters frozen at zero (or chosen only on a held-out validation split) and compare against PiSA-SR and OSEDiff on LPIPS, DISTS, and FID; if the margins shrink to noise or reverse — for instance if DRealSR LPIPS climbs back above 0.30 or FID above 135 — the reported balance of perceptual quality and fidelity is an artifact of test-set tuning rather than of the trained network.

Watch

Extended reading notes

Core claim

The central claim, stated on the paper's own terms, is that a one-step latent diffusion model can restore images with both high perceptual quality and faithful content — the combination prior one-step methods sacrificed. The paper locates the failure in the residual latent formulation z0 = zL − r: a single globally predicted residual is unstable and leaves high-frequency noise uncorrected, while the latent encoder discards low-frequency structure. FiDeSR's answer is three mechanisms acting at three stages. During training, Detail-aware Weighting (DAW) builds a per-pixel difficulty map W = D ⊙ E from a spatial detail map (Sobel, Laplacian, local variance of the ground truth) times an error ma

Load-bearing premise

The headline numbers assume it is legitimate to tune the inference-time injection strengths (lf_alpha, hf_beta) on the RealSR evaluation set and then report that same set's improved scores; if the working point must be fixed without seeing test data, the claimed advantage over the next-best one-step method is optimistic.

Editorial extensions

If this is right

  • If FiDeSR holds up, real-world super-resolution no longer needs multi-step denoising for fidelity: its single-step output beats 20–200-step methods on perceptual metrics while running roughly 25–100 times faster.
  • The LFIM injection strengths (lf_alpha, hf_beta) provide a no-retraining control dial between distortion-oriented and perception-oriented outputs, so one model can serve both archival-fidelity and display-sharpness use cases.
  • The measured reduction in high-frequency noise prediction error (1.24–1.99% across datasets) indicates LRRB addresses a general weakness of one-step distillation rather than a data-specific artifact, and could be reused in other single-step generative restoration pipelines.
  • The FID gains and user-study votes imply restored images sit closer to the true image distribution as judged both by statistics and by human preference.

Reading between the lines

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

  • Editorial: the LFIM working point (lf_alpha=0.2, hf_beta=0.2) was selected by scanning injection strengths on the RealSR test set, and that same set produces the headline RealSR row in Table 1; a skeptic should read those numbers as the ceiling of a tuned module, not of the trained network alone.
  • Editorial: if the diagnosis is right — latent compression causes low-frequency drift and one-step truncation causes high-frequency loss — then LFIM-style frequency reinjection could serve as a drop-in inference module for other latent-diffusion restoration tasks, such as deblurring or face restoration, regardless of training objective.
  • Editorial: a natural testable extension is to replace the global constants α and β with per-image or per-patch selection, e.g., optimizing them against a no-reference quality score; the reported monotonicity of PSNR/SSIM in α and of MUSIQ/MANIQA in β suggests such an autotuner would stay on a smooth surface.
  • Editorial: because the DAW error map uses an LPIPS backbone, the reported fidelity gains may inherit that metric's bias; retraining with a different perceptual error map is a direct robustness check the paper does not run.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes FiDeSR, a one-step latent diffusion super-resolution framework that combines three components: a Detail-aware Weighting (DAW) strategy that reweights training losses by a spatial difficulty map, a Latent Residual Refinement Block (LRRB) that refines the predicted residual, and a Latent Frequency Injection Module (LFIM) that injects low- and high-frequency components at inference. The authors claim that FiDeSR achieves a better perceptual-fidelity balance than existing one-step and even multi-step diffusion SR methods, with strong results on RealSR, DRealSR, and a DIV2K-based synthetic test set, while requiring only one diffusion step.

Significance. If the empirical claims survive independent validation, this is a useful contribution to efficient real-world SR: it suggests that a one-step latent diffusion model can improve both perceptual quality and structural fidelity through loss reweighting, residual refinement, and tunable frequency injection. The ablations in Table 4 show directionally sensible and monotonic effects of LF/HF injection, and the LRRB/DAW ablations in Table 2 indicate that the training-side components help on no-reference metrics. The paper is generally well structured and the proposed modules are clearly described. However, the headline comparisons rest on an evaluation-protocol concern (inference-time parameters selected on the test set) and on a quantitative inconsistency in Table 3 that need to be resolved before the central claim can be accepted.

major comments (3)
  1. [Sec. 4.2 (Table 1) and Supp. Sec. G (Table 8)] The LFIM injection strengths lf_alpha=0.2 and hf_beta=0.2 were selected by scanning on the RealSR test set, and the same RealSR test set is then reported as the headline row in Table 1. This makes the RealSR comparison a test-set-tuned result, and the DRealSR and DIV2K rows inherit hyperparameters chosen with knowledge of a same-domain test set. The reported advantages over PiSA-SR on RealSR (e.g., FID 109.68 vs 124.18) may therefore be partly an artifact of tuning rather than a property of the one-step model. Please re-select the injection strengths on a held-out validation split (or pre-specify them before evaluation) and re-run all benchmark rows; if the parameters are meant to be user-adjustable controls, the state-of-the-art claim should be qualified accordingly.
  2. [Sec. 4.3, Table 3] The RealSR row is internally inconsistent: the baseline MSE is 0.1029 and the LRRB MSE is 0.1045, i.e., LRRB increases the error, yet the table reports a 1.62% improvement. The average row also does not match the dataset rows (0.1049 vs 0.1032). This contradicts the claim in Sec. 4.3 that the LRRB-equipped model consistently reduces the high-frequency noise prediction error. Because Table 3 is the quantitative evidence for the LRRB component, this inconsistency is load-bearing and must be corrected and independently verified.
  3. [Eqs. (5), (6), (10) and Table 2] DAW constructs its perceptual error map using LPIPS, and the training loss includes a spatially weighted LPIPS term; LPIPS is also a headline evaluation metric. The paper should clarify the degree to which the reported LPIPS gains reflect direct per-pixel optimization rather than generalizable perceptual improvement. Currently Table 2 reports only no-reference metrics for the DAW ablation and therefore does not disentangle this. Please add an ablation with p=0 (or unweighted LPIPS) and report the full metric set, including LPIPS, DISTS, and FID, to demonstrate that the gains are not solely due to optimizing the evaluation metric.
minor comments (5)
  1. [Sec. 4.1] The training set includes DIV2K and the synthetic test set is cropped from DIV2K-validation. Please state explicitly that the training split excludes the validation images used for testing, to rule out train/test overlap.
  2. [Alg. 1 and Sec. 3.3] The coefficients p, w_max, and alpha appear in the DAW implementation but are not given values in the main text. Report the chosen values and, if possible, a sensitivity study.
  3. [Eq. (3)] The detail map D is defined as the mean response of Sobel, Laplacian, and variance filters, but the three operators produce outputs on very different scales. Clarify whether each operator is normalized before averaging.
  4. [Table 6] The caption and table use 'PISA-SR' while the rest of the paper uses 'PiSA-SR'. Please standardize the name.
  5. [Sec. 3.6] The notation Δ_LP and Δ_HP is introduced without formal definition. Define these quantities explicitly and specify how the Butterworth filters and the spatial/channel gates are computed.

Circularity Check

1 steps flagged · score 6.0 of 10

RealSR headline row is a test-set-tuned LFIM working point, so part of the claimed SOTA balance is fitted rather than predicted.

  1. fitted input called prediction [Supplementary Material Sec. G (Table 8); Table 1 RealSR row]
    "Supp. Sec. G: 'In our implementation, we adopt a balanced configuration of lf_alpha=0.2 and hf_beta=0.2, which we found to yield a well-rounded compromise...' Table 8 caption: 'Ablation study on different LF/HF injection strength ratios of LFIM(lf alpha,hf beta) on the RealSR dataset.' Table 1 RealSR row: 'FiDeSR-1s 26.02 0.7457 0.2626 0.1965 0.6896 5.3194 69.8245 0.6681 109.68'."

    The Table 1 RealSR row is exactly the (0.2,0.2) working point reported in Table 8, and the text says this configuration was found by scanning LF/HF injection strengths on the RealSR dataset. Since RealSR is the evaluation set for that row, the reported PSNR/LPIPS/FID/etc. are a test-set-tuned choice rather than a prediction of the trained model at a pre-specified setting. The improvement over the no-LFIM baseline and over competitors on RealSR is therefore partly an artifact of parameter selection on the test set. The DRealSR and DIV2K rows inherit the same (0.2,0.2) choice, which was made with knowledge of a same-domain real-world test set. This is the fitted-input-called-prediction pattern.

full rationale

The only load-bearing circular step I can exhibit is the LFIM test-set tuning. The training-side components (DAW, LRRB) are described by explicit equations (Eqs. 3-12) and are not defined in terms of the reported benchmark numbers; there is no self-citation chain or imported uniqueness theorem. The use of LPIPS both as a weighted training loss and as a reported metric is standard supervised SR practice and, by itself, does not make the comparison circular; it is a correctness concern only if one treats LPIPS as an independent perceptual oracle. I did not count the Table 3 RealSR entry (0.1045 vs 0.1029 listed as 1.62% improvement) as circularity: it is an internal arithmetic inconsistency that signals the reported numbers need independent verification, but it does not make the derivation equivalent to its inputs. Because the headline RealSR row and the cross-dataset configuration inherit a test-set-tuned working point, the central empirical claim is partially fitted; hence score 6 rather than 0-2.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

No new physical or conceptual entities (particles, forces, dimensions, conserved quantities) are introduced. The DAW, LRRB, and LFIM modules are architectural components proposed as the method itself; their effects are assessed only inside the paper, and the tunable LFIM intensities are logged as free parameters above rather than as invented entities.

free parameters (7)
  • hf_beta (HF injection strength) = 0.2
    Inference-time high-frequency injection strength; scanned over 0.1-0.5 on the RealSR test set (Tables 4/8) and set to 0.2 for all headline results (Supp. Sec. G).
  • lf_alpha (LF injection strength) = 0.2
    Inference-time low-frequency injection strength; scanned on RealSR (Tables 4/8); the 0.2/0.2 combination is used in Table 1 and described as 'a well-rounded compromise' found by the authors.
  • DAW perceptual mix coefficient p = not specified
    Blends L1 and LPIPS error maps in Eq. 6 / Alg. 1 line 7; no value is given, so the exact training objective is not reproducible from the text.
  • DAW weight normalization (w_max, alpha) = not specified
    Alg. 1 lines 9-10 scale W via tanh(blur(D.E)/w_max).w_max and mean-normalize 1+alpha.W; both values are omitted.
  • Loss weights lambda_mse, lambda_lpips, lambda_reg = 1, 2, unspecified
    Eq. 10/11: lambda_mse=1 and lambda_lpips=2 are stated in Sec. 4.1, but lambda_reg is never given a numeric value, affecting the CSD term's contribution.
  • LoRA rank = 8
    Selected after an ablation (Table 7) showing ranks 4/8/16 give similar performance; a mild free choice with documented insensitivity.
  • HF error radial cutoff r_c = 0.8
    Defines 'high-frequency noise prediction error' in Table 3 (Supp. Sec. E.1, top 20% frequencies); the reported 1.62% LRRB improvement is relative to this arbitrary cutoff.
assumptions (5)
  • domain assumption Frozen SD2.1-base VAE and U-Net plus trainable LoRA provide a valid prior for expressing the LQ-to-HQ latent correction
    Sec. 4.1: with the backbone frozen and only LoRA and LRRB trained, the model is assumed capable of representing the residual z_L -> z_H; no capacity or distortion analysis is given.
  • domain assumption CSD (variational score distillation) provides a stable and sufficient one-step regularization signal
    Sec. 3.5, Eq. 11: the L_reg term is inherited from PiSA-SR/ProlificDreamer without re-derivation; the stability of one-step CSD at this scale is taken for granted.
  • domain assumption Real-ESRGAN degradation produces training pairs that transfer to RealSR/DRealSR real-world degradations
    Sec. 4.1: all training pairs are synthesized with the Real-ESRGAN pipeline; the 'real-world superiority' claim depends on this transfer holding.
  • domain assumption Butterworth LF/HF decomposition of SD2.1 VAE latents corresponds to perceptually meaningful image frequencies
    Sec. 3.6 and Supp. Sec. G: injection behavior is validated only through in-paper ablations; no external evidence links latent-space frequency components to pixel-space detail/structure.
  • domain assumption LPIPS is a faithful perceptual error measure in the DAW error map and as an evaluation metric
    Eq. 5 and Table 1: LPIPS simultaneously defines the training weighting and the reported perceptual outcome; its validity as ground truth is assumed rather than argued.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FiDeSR: High-Fidelity and Detail-Preserving One-Step Diffusion Super-Resolution." pith.science (2026). https://pith.science/paper/L5K4MKKI

@misc{pith2026260302692,
  author       = {Pith},
  title        = {Pith review of: FiDeSR: High-Fidelity and Detail-Preserving One-Step Diffusion Super-Resolution},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/L5K4MKKI}},
  note         = {Machine review of arXiv:2603.02692}
}
read the original abstract

Diffusion-based approaches have recently driven remarkable progress in real-world image super-resolution (SR). However, existing methods still struggle to simultaneously preserve fine details and ensure high-fidelity reconstruction, often resulting in suboptimal visual quality. In this paper, we propose FiDeSR, a high-fidelity and detail-preserving one-step diffusion super-resolution framework. During training, we introduce a detail-aware weighting strategy that adaptively emphasizes regions where the model exhibits higher prediction errors. During inference, low- and high-frequency adaptive enhancers further refine the reconstruction without requiring model retraining, enabling flexible enhancement control. To further improve the reconstruction accuracy, FiDeSR incorporates a residual-in-residual noise refinement, which corrects prediction errors in the diffusion noise and enhances fine detail recovery. FiDeSR achieves superior real-world SR performance compared to existing diffusion-based methods, producing outputs with both high perceptual quality and faithful content restoration. The source code will be released at: https://github.com/Ar0Kim/FiDeSR.

Figures

Figures reproduced from arXiv: 2603.02692 by the authors.

Figure 1
Figure 1. Performance comparison among Real-ISR methods on three perceptual–fidelity metric pairs: PSNR vs. MANIQA (left), SSIM [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Example failure cases of diffusion-based Real-ISR methods. (b) AddSR introduces structural distortion and low-frequency [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Overall framework of FiDeSR. (a) Training process: LQ image [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Comparison of the pipeline among ESRGAN, PiSA-SR, [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Qualitative comparisons with state-of-the-art DM-based SR methods. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Ground-truth images used in the user study along with the voting results for their corresponding restored outputs. [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: Qualitative comparisons between FiDeSR and GAN-based Real-ISR methods. [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]
Figure 8
Figure 8. Figure 8: Qualitative comparisons between FiDeSR and different diffusion-based methods on DIV2K dataset. FiDeSR effectively recon [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 9
Figure 9. Figure 9: Qualitative comparisons between FiDeSR and different diffusion-based methods on DRealSR and RealSR dataset. FiDeSR [PITH_FULL_IMAGE:figures/full_fig_p016_9.png]
Figure 10
Figure 10. Figure 10: Spatial distribution of high-frequency noise prediction [PITH_FULL_IMAGE:figures/full_fig_p017_10.png]
Figure 11
Figure 11. Figure 11: Visualization of the Detail-aware Weighting (DAW) [PITH_FULL_IMAGE:figures/full_fig_p017_11.png]
Figure 12
Figure 12. Figure 12: Qualitative ablation study illustrating the contributions of the LRRB and DAW modules. Even under challenging degradation [PITH_FULL_IMAGE:figures/full_fig_p018_12.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

62 extracted references · 5 linked inside Pith

  1. [1]

    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. InPro- ceedings of the IEEE conference on computer vision and pat- tern recognition workshops, pages 126–135, 2017. 6, 12

  2. [2]

    Guidesr: Rethinking guidance for one-step high-fidelity diffusion-based super-resolution

    Aditya Arora, Zhengzhong Tu, Yufei Wang, Ruizheng Bai, Jian Wang, and Sizhuo Ma. Guidesr: Rethinking guidance for one-step high-fidelity diffusion-based super-resolution. InProceedings of the IEEE/CVF International Conference on Computer Vision Workshops (ICCVW), 2025. Accepted. 2, 3

  3. [3]

    Toward real-world single image super-resolution: A new benchmark and a new model

    Jianrui Cai, Hui Zeng, Hongwei Yong, Zisheng Cao, and Lei Zhang. Toward real-world single image super-resolution: A new benchmark and a new model. InProceedings of the IEEE/CVF international conference on computer vision, pages 3086–3095, 2019. 6, 12

  4. [4]

    Swinfsr: Stereo image super-resolution using swinir and frequency domain knowledge

    Ke Chen, Liangyan Li, Huan Liu, Yunzhe Li, Congling Tang, and Jun Chen. Swinfsr: Stereo image super-resolution using swinir and frequency domain knowledge. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 1764–1774, 2023. 3

  5. [5]

    Freqformer: Frequency-aware transformer for lightweight image super-resolution

    Tao Dai, Jianping Wang, Hang Guo, Jinmin Li, Jinbao Wang, and Zexuan Zhu. Freqformer: Frequency-aware transformer for lightweight image super-resolution. InProceedings of the International Joint Conference on Artificial Intelligence, pages 731–739, 2024. 3

  6. [6]

    Diffusion models beat gans on image synthesis.Advances in neural informa- tion processing systems, 34:8780–8794, 2021

    Prafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis.Advances in neural informa- tion processing systems, 34:8780–8794, 2021. 1

  7. [7]

    Image quality assessment: Unifying structure and texture similarity.IEEE transactions on pattern analysis and ma- chine intelligence, 44(5):2567–2581, 2020

    Keyan Ding, Kede Ma, Shiqi Wang, and Eero P Simoncelli. Image quality assessment: Unifying structure and texture similarity.IEEE transactions on pattern analysis and ma- chine intelligence, 44(5):2567–2581, 2020. 6

  8. [8]

    Tsd-sr: One-step diffusion with target score distillation for real-world image super-resolution

    Linwei Dong, Qingnan Fan, Yihong Guo, Zhonghao Wang, Qi Zhang, Jinwei Chen, Yawei Luo, and Changqing Zou. Tsd-sr: One-step diffusion with target score distillation for real-world image super-resolution. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 23174–23184, 2025. 2, 3

Show all 62 references
  1. [9]

    Scaling recti- fied flow transformers for high-resolution image synthesis

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas M ¨uller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, et al. Scaling recti- fied flow transformers for high-resolution image synthesis. InForty-first international conference on machi...

  2. [10]

    Generative adversarial nets.Advances in neural information processing systems, 27, 2014

    Ian J Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets.Advances in neural information processing systems, 27, 2014. 1

  3. [11]

    Deep wavelet prediction for image super- resolution

    Tiantong Guo, Hojjat Seyed Mousavi, Tiep Huu Vu, and Vishal Monga. Deep wavelet prediction for image super- resolution. InProceedings of the IEEE conference on com- puter vision and pattern recognition workshops, pages 104– 113, 2017. 3

  4. [12]

    Gans trained by a two time-scale update rule converge to a local nash equilib- rium.Advances in neural information processing systems, 30, 2017

    Martin Heusel, Hubert Ramsauer, Thomas Unterthiner, Bernhard Nessler, and Sepp Hochreiter. Gans trained by a two time-scale update rule converge to a local nash equilib- rium.Advances in neural information processing systems, 30, 2017. 6

  5. [13]

    Denoising dif- fusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising dif- fusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020. 1, 3

  6. [14]

    Parameter-efficient transfer learning for nlp

    Neil Houlsby, Andrei Giurgiu, Stanislaw Jastrzebski, Bruna Morrone, Quentin De Laroussilhe, Andrea Gesmundo, Mona Attariyan, and Sylvain Gelly. Parameter-efficient transfer learning for nlp. InInternational conference on machine learning, pages 2790–2799. PMLR, 2019. 3

  7. [15]

    Lora: Low-rank adaptation of large language models.ICLR, 1(2):3, 2022

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen- Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, et al. Lora: Low-rank adaptation of large language models.ICLR, 1(2):3, 2022. 3

  8. [16]

    Focal frequency loss for image reconstruction and synthesis

    Liming Jiang, Bo Dai, Wayne Wu, and Chen Change Loy. Focal frequency loss for image reconstruction and synthesis. InProceedings of the IEEE/CVF international conference on computer vision, pages 13919–13929, 2021. 3

  9. [17]

    A style-based generator architecture for generative adversarial networks

    Tero Karras, Samuli Laine, and Timo Aila. A style-based generator architecture for generative adversarial networks. InProceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 4401–4410, 2019. 6

  10. [18]

    Musiq: Multi-scale image quality transformer

    Junjie Ke, Qifei Wang, Yilin Wang, Peyman Milanfar, and Feng Yang. Musiq: Multi-scale image quality transformer. InProceedings of the IEEE/CVF international conference on computer vision, pages 5148–5157, 2021. 6

  11. [19]

    Deep laplacian pyramid networks for fast and accurate super-resolution

    Wei-Sheng Lai, Jia-Bin Huang, Narendra Ahuja, and Ming- Hsuan Yang. Deep laplacian pyramid networks for fast and accurate super-resolution. InProceedings of the IEEE con- ference on computer vision and pattern recognition, pages 624–632, 2017. 3

  12. [20]

    Photo- realistic single image super-resolution using a generative ad- versarial network

    Christian Ledig, Lucas Theis, Ferenc Husz´ar, Jose Caballero, Andrew Cunningham, Alejandro Acosta, Andrew Aitken, Alykhan Tejani, Johannes Totz, Zehan Wang, et al. Photo- realistic single image super-resolution using a generative ad- versarial network. InProceedings of the IEE...

  13. [21]

    Srdiff: Single image super-resolution with diffusion probabilistic models

    Haoying Li, Yifan Yang, Meng Chang, Shiqi Chen, Huajun Feng, Zhihai Xu, Qi Li, and Yueting Chen. Srdiff: Single image super-resolution with diffusion probabilistic models. Neurocomputing, 479:47–59, 2022. 3 9

  14. [22]

    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 Deman- dolx, et al. Lsdir: A large scale dataset for image restoration. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1775–178...

  15. [23]

    A timestep-adaptive frequency-enhancement framework for diffusion-based im- age super-resolution

    Yueying Li, Hanbin Zhao, Jiaqing Zhou, Guozhi Xu, Tian- lei Hu, Gang Chen, and Haobo Wang. A timestep-adaptive frequency-enhancement framework for diffusion-based im- age super-resolution. InProceedings of the Thirty-Fourth In- ternational Joint Conference on Artificial Intell...

  16. [24]

    Details or artifacts: A locally discriminative learning approach to realistic im- age super-resolution

    Jie Liang, Hui Zeng, and Lei Zhang. Details or artifacts: A locally discriminative learning approach to realistic im- age super-resolution. InProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pages 5657–5666, 2022. 12

  17. [25]

    Diff- bir: Toward blind image restoration with generative diffusion prior

    Xinqi Lin, Jingwen He, Ziyan Chen, Zhaoyang Lyu, Bo Dai, Fanghua Yu, Yu Qiao, Wanli Ouyang, and Chao Dong. Diff- bir: Toward blind image restoration with generative diffusion prior. InEuropean conference on computer vision, pages 430–448. Springer, 2024. 3

  18. [26]

    Decoupled weight decay regularization.arXiv preprint arXiv:1711.05101, 2017

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization.arXiv preprint arXiv:1711.05101, 2017. 6

  19. [27]

    Waving goodbye to low-res: A diffusion-wavelet approach for image super-resolution

    Brian B Moser, Stanislav Frolov, Federico Raue, Sebastian Palacio, and Andreas Dengel. Waving goodbye to low-res: A diffusion-wavelet approach for image super-resolution. In 2024 International Joint Conference on Neural Networks (IJCNN), pages 1–8. IEEE, 2024. 3

  20. [28]

    Fcanet: Frequency channel attention networks

    Zequn Qin, Pengyi Zhang, Fei Wu, and Xi Li. Fcanet: Frequency channel attention networks. InProceedings of the IEEE/CVF international conference on computer vision, pages 783–792, 2021. 3

  21. [29]

    Xpsr: Cross-modal priors for diffusion-based image super-resolution

    Yunpeng Qu, Kun Yuan, Kai Zhao, Qizhi Xie, Jinhua Hao, Ming Sun, and Chao Zhou. Xpsr: Cross-modal priors for diffusion-based image super-resolution. InEuropean Con- ference on Computer Vision, pages 285–303. Springer, 2024. 3

  22. [30]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High-resolution image synthesis with latent diffusion models. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10684–10695, 2022. 2, 3

  23. [31]

    Image super- resolution via iterative refinement.IEEE transactions on pattern analysis and machine intelligence, 45(4):4713–4726,

    Chitwan Saharia, Jonathan Ho, William Chan, Tim Sali- mans, David J Fleet, and Mohammad Norouzi. Image super- resolution via iterative refinement.IEEE transactions on pattern analysis and machine intelligence, 45(4):4713–4726,

  24. [32]

    Adversarial diffusion distillation

    Axel Sauer, Dominik Lorenz, Andreas Blattmann, and Robin Rombach. Adversarial diffusion distillation. InEuropean Conference on Computer Vision, pages 87–103. Springer,

  25. [33]

    Resdiff: Combining cnn and diffusion model for image super-resolution

    Shuyao Shang, Zhengyang Shan, Guangxing Liu, LunQian Wang, XingHua Wang, Zekai Zhang, and Jinglin Zhang. Resdiff: Combining cnn and diffusion model for image super-resolution. InProceedings of the AAAI Conference on Artificial Intelligence, pages 8975–8983, 2024. 3

  26. [34]

    Multi-scale adversarial diffusion network for image super- resolution.Scientific Reports, 15(1):11690, 2025

    Yanli Shi, Xianhe Zhang, Yi Jia, and Jinxing Zhao. Multi-scale adversarial diffusion network for image super- resolution.Scientific Reports, 15(1):11690, 2025. 3

  27. [35]

    Freeu: Free lunch in diffusion u-net

    Chenyang Si, Ziqi Huang, Yuming Jiang, and Ziwei Liu. Freeu: Free lunch in diffusion u-net. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4733–4743, 2024. 3

  28. [36]

    Score-based generative modeling through stochastic differential equa- tions.arXiv preprint arXiv:2011.13456, 2020

    Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Ab- hishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equa- tions.arXiv preprint arXiv:2011.13456, 2020. 1

  29. [37]

    Stable diffusion.https://stability

    Stability.ai. Stable diffusion.https://stability. ai/stable-diffusion. 6

  30. [38]

    Improving the stability and efficiency of diffusion models for content consistent super- resolution.arXiv preprint arXiv:2401.00877, 2023

    Lingchen Sun, Rongyuan Wu, Jie Liang, Zhengqiang Zhang, Hongwei Yong, and Lei Zhang. Improving the stability and efficiency of diffusion models for content consistent super- resolution.arXiv preprint arXiv:2401.00877, 2023. 2

  31. [39]

    Pixel-level and semantic-level ad- justable super-resolution: A dual-lora approach

    Lingchen Sun, Rongyuan Wu, Zhiyuan Ma, Shuaizheng Liu, Qiaosi Yi, and Lei Zhang. Pixel-level and semantic-level ad- justable super-resolution: A dual-lora approach. InProceed- ings of the Computer Vision and Pattern Recognition Con- ference, pages 2333–2343, 2025. 2, 3, 5

  32. [40]

    Ntire 2017 challenge on single image super-resolution: Methods and results

    Radu Timofte, Eirikur Agustsson, Luc Van Gool, Ming- Hsuan Yang, and Lei Zhang. Ntire 2017 challenge on single image super-resolution: Methods and results. InProceed- ings of the IEEE conference on computer vision and pattern recognition workshops, pages 114–125, 2017. 6

  33. [41]

    Ex- ploring clip for assessing the look and feel of images

    Jianyi Wang, Kelvin CK Chan, and Chen Change Loy. Ex- ploring clip for assessing the look and feel of images. InPro- ceedings of the AAAI conference on artificial intelligence, pages 2555–2563, 2023. 6

  34. [42]

    Exploiting diffusion prior for real-world image super-resolution.International Journal of Computer Vision, 132(12):5929–5949, 2024

    Jianyi Wang, Zongsheng Yue, Shangchen Zhou, Kelvin CK Chan, and Chen Change Loy. Exploiting diffusion prior for real-world image super-resolution.International Journal of Computer Vision, 132(12):5929–5949, 2024. 2, 3

  35. [43]

    Esrgan: En- hanced super-resolution generative adversarial networks

    Xintao Wang, Ke Yu, Shixiang Wu, Jinjin Gu, Yihao Liu, Chao Dong, Yu Qiao, and Chen Change Loy. Esrgan: En- hanced super-resolution generative adversarial networks. In Proceedings of the European conference on computer vision (ECCV) workshops, pages 0–0, 2018. 1, 5

  36. [44]

    Real-esrgan: Training real-world blind super-resolution with pure synthetic data

    Xintao Wang, Liangbin Xie, Chao Dong, and Ying Shan. Real-esrgan: Training real-world blind super-resolution with pure synthetic data. InProceedings of the IEEE/CVF inter- national conference on computer vision, pages 1905–1914,

  37. [45]

    Frequency- domain refinement with multiscale diffusion for super res- olution.arXiv preprint arXiv:2405.10014, 2024

    Xingjian Wang, Li Chai, and Jiming Chen. Frequency- domain refinement with multiscale diffusion for super res- olution.arXiv preprint arXiv:2405.10014, 2024. 3

  38. [46]

    Reconstruct-and-generate diffusion model for detail-preserving image denoising.arXiv preprint arXiv:2309.10714, 2023

    Yujin Wang, Lingen Li, Tianfan Xue, and Jinwei Gu. Reconstruct-and-generate diffusion model for detail-preserving image denoising.arXiv preprint arXiv:2309.10714, 2023. 3

  39. [47]

    Sinsr: diffusion-based image super- resolution in a single step

    Yufei Wang, Wenhan Yang, Xinyuan Chen, Yaohui Wang, Lanqing Guo, Lap-Pui Chau, Ziwei Liu, Yu Qiao, Alex C Kot, and Bihan Wen. Sinsr: diffusion-based image super- resolution in a single step. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, p...

  40. [48]

    Image quality assessment: from error visibility to structural similarity.IEEE transactions on image processing, 13(4):600–612, 2004

    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. 6

  41. [49]

    Prolificdreamer: High-fidelity and diverse text-to-3d generation with variational score distilla- tion.Advances in neural information processing systems, 36: 8406–8441, 2023

    Zhengyi Wang, Cheng Lu, Yikai Wang, Fan Bao, Chongxuan Li, Hang Su, and Jun Zhu. Prolificdreamer: High-fidelity and diverse text-to-3d generation with variational score distilla- tion.Advances in neural information processing systems, 36: 8406–8441, 2023. 3, 5

  42. [50]

    Component divide-and-conquer for real-world image super-resolution

    Pengxu Wei, Ziwei Xie, Hannan Lu, Zongyuan Zhan, Qix- iang Ye, Wangmeng Zuo, and Liang Lin. Component divide-and-conquer for real-world image super-resolution. In European conference on computer vision, pages 101–117. Springer, 2020. 6, 12

  43. [51]

    One-step effective diffusion network for real-world image super-resolution.Advances in Neural Information Process- ing Systems, 37:92529–92553, 2024

    Rongyuan Wu, Lingchen Sun, Zhiyuan Ma, and Lei Zhang. One-step effective diffusion network for real-world image super-resolution.Advances in Neural Information Process- ing Systems, 37:92529–92553, 2024. 2, 3

  44. [52]

    Seesr: Towards semantics- aware real-world image super-resolution

    Rongyuan Wu, Tao Yang, Lingchen Sun, Zhengqiang Zhang, Shuai Li, and Lei Zhang. Seesr: Towards semantics- aware real-world image super-resolution. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 25456–25467, 2024. 3

  45. [53]

    Addsr: Accelerating diffusion- based blind super-resolution with adversarial diffusion dis- tillation.arXiv preprint arXiv:2404.01717, 2024

    Rui Xie, Chen Zhao, Kai Zhang, Zhenyu Zhang, Jun Zhou, Jian Yang, and Ying Tai. Addsr: Accelerating diffusion- based blind super-resolution with adversarial diffusion dis- tillation.arXiv preprint arXiv:2404.01717, 2024. 3

  46. [54]

    Image super-resolution via sparse representation.IEEE transactions on image processing, 19(11):2861–2873, 2010

    Jianchao Yang, John Wright, Thomas S Huang, and Yi Ma. Image super-resolution via sparse representation.IEEE transactions on image processing, 19(11):2861–2873, 2010. 1

  47. [55]

    Maniqa: Multi-dimension attention network for no-reference image quality assessment

    Sidi Yang, Tianhe Wu, Shuwei Shi, Shanshan Lao, Yuan Gong, Mingdeng Cao, Jiahao Wang, and Yujiu Yang. Maniqa: Multi-dimension attention network for no-reference image quality assessment. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages...

  48. [56]

    Pixel-aware stable diffusion for realistic image super-resolution and personalized stylization

    Tao Yang, Rongyuan Wu, Peiran Ren, Xuansong Xie, and Lei Zhang. Pixel-aware stable diffusion for realistic image super-resolution and personalized stylization. InEuropean conference on computer vision, pages 74–91. Springer, 2024. 3

  49. [57]

    One-step diffusion with distribution matching distillation

    Tianwei Yin, Micha ¨el Gharbi, Richard Zhang, Eli Shecht- man, Fredo Durand, William T Freeman, and Taesung Park. One-step diffusion with distribution matching distillation. In Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 6613–6623...

  50. [58]

    Resshift: Efficient diffusion model for image super- resolution by residual shifting.Advances in Neural Infor- mation Processing Systems, 36:13294–13307, 2023

    Zongsheng Yue, Jianyi Wang, and Chen Change Loy. Resshift: Efficient diffusion model for image super- resolution by residual shifting.Advances in Neural Infor- mation Processing Systems, 36:13294–13307, 2023. 2, 3

  51. [59]

    Designing a practical degradation model for deep blind im- age super-resolution

    Kai Zhang, Jingyun Liang, Luc Van Gool, and Radu Timofte. Designing a practical degradation model for deep blind im- age super-resolution. InProceedings of the IEEE/CVF inter- national conference on computer vision, pages 4791–4800,

  52. [60]

    A feature-enriched completely blind image quality evaluator.IEEE Transactions on Image Processing, 24(8):2579–2591, 2015

    Lin Zhang, Lei Zhang, and Alan C Bovik. A feature-enriched completely blind image quality evaluator.IEEE Transactions on Image Processing, 24(8):2579–2591, 2015. 6

  53. [61]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shecht- man, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. InProceedings of the IEEE conference on computer vision and pattern recogni- tion, pages 586–595, 2018. 6

  54. [62]

    Recognize anything: A strong image tagging model

    Youcai Zhang, Xinyu Huang, Jinyu Ma, Zhaoyang Li, Zhaochuan Luo, Yanchun Xie, Yuzhuo Qin, Tong Luo, Yaqian Li, Shilong Liu, et al. Recognize anything: A strong image tagging model. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 172...

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.