Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

Arbitrary-steps Image Super-resolution via Diffusion Inversion

T0 review · 4 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read This paper claims that image super-resolution can be done from a frozen diffusion model by training a small noise predictor to construct a single intermediate starting state, enabling one-to-five-step sampling with single-step quality…

desk verdict A genuinely practical one-to-five-step diffusion SR method with solid held-out benchmarks and one load-bearing clarity issue that should be fixed, not fatal. read the letter →

arxiv 2412.09013 v2 pith:HHALNZJJ submitted 2024-12-12 cs.CV

classification cs.CV
keywords diffusioninversionimagesuper-resolutionpartialnoisepredictionpredictorarbitrary-stepsamplingStableblindone-step
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

InvSR claims that image super-resolution can be reduced to choosing one good noise map. Instead of running a large diffusion model from pure noise or fine-tuning it, the method trains a compact 33.84M-parameter noise predictor that, given a low-resolution image and a starting timestep, produces the noise needed to build an intermediate diffusion state. From that state, a completely frozen Stable Diffusion Turbo model generates the high-resolution image in one to five sampling steps. The paper reports that even a single step beats or matches recent dedicated one-step diffusion SR methods across perceptual and distortion metrics, while allowing the user to trade fidelity for realism by changing the starting timestep. If this holds, the practical cost of diffusion-based SR drops sharply while keeping the large pretrained model untouched.

What carries the argument

The load-bearing object is the Partial Noise Prediction starting-state formula in Eq. (8): $x_{\kappa_M} = \sqrt{\bar\alpha_{\kappa_M}}\,y_0 + \sqrt{1-\bar\alpha_{\kappa_M}}\,f_w(y_0, \kappa_M)$. It collapses the full inversion problem, which would require predicting $T$ noise maps, into a single noise map for one chosen intermediate timestep, and it is the only place where the low-resolution image enters the diffusion trajectory. The noise predictor $f_w$, built on a VQGAN encoder with two downsampling blocks and self-attention, is trained to output Gaussian mean and variance parameters via the reparameterization trick, and its time embedding is what makes the starting timestep selectable at inference. Intermediate sampling steps are then ordinary DDIM/DDPM updates with random noise, relying on the high signal-to-noise ratio of timesteps at or below 250 to keep the frozen SD-Turbo denoiser on track.

What would settle it

A direct falsifier is a controlled swap: in Eq. (8) at the trained starting steps, replace the predictor's output with zero-mean Gaussian noise of the same standard deviation, keep the frozen SD-Turbo sampler unchanged, and evaluate on ImageNet-Test; if one-step PSNR and LPIPS do not clearly worsen, the learned noise map is not what carries the super-resolution result, and a reproduction that fails to match Table 2's one-step numbers would likewise refute the headline claim.

Watch

Extended reading notes

Core claim

The central claim is that the reverse diffusion process for super-resolution does not need per-step noise optimization or backbone fine-tuning. The paper's construction starts from the DDPM marginal $x_\tau = \sqrt{\bar\alpha_\tau}\,x_0 + \sqrt{1-\bar\alpha_\tau}\,\xi$ and replaces the unknown high-resolution image $x_0$ with the given low-resolution image $y_0$, using a trained network $f_w$ to supply the noise: $x_\tau = \sqrt{\bar\alpha_\tau}\,y_0 + \sqrt{1-\bar\alpha_\tau}\,f_w(y_0, \tau)$. The predictor is time-conditioned and trained with L2, LPIPS, and GAN losses in the latent space of SD-Turbo, with starting timesteps 250, 200, 150, and 100; at inference the user picks a starting step and runs one to five DDIM-style steps. The paper's evidence is that InvSR-1 reaches 24.14 dB PSNR and 0.2517 LPIPS on ImageNet-Test, outperforming the one-step OSEDiff baseline on all seven reported metrics, and that the predicted noise map is visibly correlated with the low-resolution image, which the authors interpret as showing that the predictor has found an LR-dependent inversion trajectory.

Load-bearing premise

The one-step result rests on the assumption that a low-resolution image perturbed by the learned noise map lands inside the narrow distribution of latent states from which the frozen denoiser can recover the high-resolution image.

Editorial extensions

If this is right

  • One-step super-resolution becomes a practical setting: InvSR-1 runs in 117 ms on an A100 and beats the dedicated one-step OSEDiff on all seven metrics reported on ImageNet-Test.
  • The number of sampling steps can be chosen per image, so noise-dominated images can use one step while blur-dominated images can use three to five steps to recover finer structure.
  • Because the diffusion backbone stays frozen, improvements in the base generative model can be inherited by retraining only the small noise predictor.
  • Reference metrics decline as sampling steps increase, so the multi-step mode is a realism tool rather than a simple accuracy booster.

Reading between the lines

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

  • An extension the paper leaves implicit is to make the starting step self-tuning: feed an automatic blur or noise estimate into the timestep selection so users do not have to choose the step count manually.
  • Since the predicted noise map is visibly correlated with the low-resolution image, the predictor can be viewed as injecting the missing high-frequency residual into the latent; the same construction could seed other conditional restoration tasks, such as deblurring or deraining, with a frozen diffusion backbone.
  • The paper does not state how the low-resolution image is encoded into SD-Turbo's latent space; a natural reading is that the same VQGAN encoder is reused, and an ablation replacing that encoder with a differently trained one would reveal how much of the super-resolution quality comes from the frozen prior versus the predictor.
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

4 major / 6 minor

Summary. The paper proposes InvSR, a super-resolution method built on diffusion inversion with a frozen SD-Turbo backbone. A noise predictor f_w is trained to map the LR observation and a timestep to the noise that should be added to the LR to form the starting latent of a short reverse sampling chain (one to five steps). Intermediate steps use randomly sampled noise, and the training objective combines L2, LPIPS, and GAN losses on latent-space estimates. Experiments on ImageNet-Test, RealSR, and RealSet80 compare the method against GAN-based and diffusion-based SR baselines, and ablations in the supplement justify the partial-noise-prediction design.

Significance. If the construction is made precise, the contribution is practically valuable: a frozen diffusion backbone plus a 33.84M-parameter noise predictor achieves competitive perceptual quality at 117 ms for one-step ×4 SR while supporting a configurable number of sampling steps. The held-out evaluation on three datasets and the public release of code and models are strengths, and the ablations in Tables 4-6 of the supplement validate the main design choices. However, the central claim rests on Eq. (8), which as written is not well-defined because the domain of y0 is never specified; this must be resolved before the empirical results can be fully assessed.

major comments (4)
  1. [Sec. 3.2.2, Eq. (8), and Algorithm 1] The domain of y0 is never specified. Section 3 defines y0 as the LR image, and Eqs. (1)-(7) are written for DDPM tensors, but the denoiser εθ is SD-Turbo, a latent diffusion model whose α-bar schedule and noise prediction operate on VQGAN latents, not on RGB images. If y0 is the pixel-space LR, then x_{κM} in Eq. (8) is an RGB tensor while εθ(x_{κM}, κM) requires a 64×64 latent for the ×4 128-to-512 task; if y0 is instead the encoded latent, the encoder is omitted and the stated f_w architecture of two downsampling blocks does not obviously produce a 64×64 noise map from a 16×16 latent. Because Eq. (8) is the sole bridge from the LR observation to the frozen diffusion model, the one-step results are not reproducible from the paper alone. Please state the exact preprocessing of y0 and the input/output shapes of f_w, or rewrite all equations consistently in latent space.
  2. [Sec. 3.2] The text states that f_w predicts the mean and variance of a Gaussian via the reparameterization trick rather than directly estimating the noise map, but Eq. (8), Eq. (10), and Algorithm 1 all use f_w(y0, t) as the noise map itself. No sampling from the predicted distribution is specified, and the variance output is never referenced in the loss or inference. Please clarify whether f_w outputs a mean that is used directly as the noise map, or add the missing sampling/reparameterization step; otherwise the training objective Eq. (11) is under-specified.
  3. [Sec. 3.2.3, Eq. (11), and training details] The paper says the whole loss is computed in the latent space and that the LPIPS loss is 'fine-tuned' in that space, but no definition is given for how LPIPS is applied to VQGAN latents or what fine-tuning means. Since LPIPS is a pixel-space perceptual metric, this sentence does not by itself define a computable loss. Provide the exact loss computation (which layers are used, what normalization is applied, and whether a separate network is trained), or move this clarification to the supplement.
  4. [Algorithm 1, line 4] The stochastic term σ_{τ_i} z_{τ_i} is never defined for the accelerated schedules used in the paper. For a diffusion step with τ_i = 250, a nonzero σ would inject substantial noise, so the reported 'one-step' result would include a stochastic perturbation at the final step, whereas the text frames one step as deterministic denoising. Specify the sampling algorithm (for example, DDIM with η = 0, or a specific σ schedule) and, if nonzero σ is used, state how the one-step numbers are obtained.
minor comments (6)
  1. [Figure 1 caption] The caption reads 'runtime (in milliseconds) highlighted by red in the sub-caption of the first example'; the phrasing is awkward and the parenthesis after 'Ours-2' is unclosed in the displayed text.
  2. [Table 1 header] The header 'Index of the sampled timesteps' should be 'Index of the sampling timesteps' for clarity.
  3. [Sec. 3.2.3] The expression f_w(y_{κ_i}, κ_i) should presumably be f_w(y_0, κ_i), since y_{κ_i} is never defined.
  4. [Eq. (2)] The product notation for \barα_t is garbled in the typeset version; it should be \barα_t = \prod_{s=1}^t α_s.
  5. [Related Work] The sentence attributing textual inversion to 'Rinon et al.' should cite 'Gal et al.' [12].
  6. [Abstract and Introduction] The phrase 'arbitrary number of sampling steps ranging from one to five' is slightly misleading; consider replacing 'arbitrary' with 'configurable' to avoid overclaiming.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the headline one-step SR results are held-out benchmark measurements of a trained composite mapping, and the self-citations are not load-bearing.

full rationale

The central claim is empirical and externally grounded. Eq. (8) constructs the starting sample from the LR and the learned noise predictor f_w, and Eq. (11) trains f_w so that the one-step denoising estimate in Eq. (9) matches the HR target; substituting Eq. (8) into Eq. (9) shows that the evaluated one-step output is exactly the composite function being optimized. That makes f_w a fitted module rather than an independent first-principles prediction, but the paper does not present it as a parameter-free derivation, and the test images are unseen during training, so the prediction is not statistically forced by construction. The self-citations (ResShift, StableSR, DifFace) appear as degradation settings, baselines, and related work and are not load-bearing premises; no uniqueness theorem is imported and no ansatz is smuggled in via citation. The main weakness is a correctness/reproducibility gap: Eq. (8) never states that y0 is VAE-encoded before being combined with f_w for SD-Turbo, which operates in VQGAN latent space. That is a domain mismatch, not a circular reduction. Accordingly, no circular step can be exhibited with a quote, and the appropriate score is 0.

Assumptions & free parameters 4 free parameters · 5 assumptions · 1 invented entities

The method adds one learned module (the noise predictor) and several manually chosen schedulers and hyperparameters. Standard DDPM machinery is imported from prior work; the only genuinely new mathematical content is the construction in Eq. (8), which is itself a trained fit rather than a derivation. No physical entities are introduced.

free parameters (4)
  • SNR threshold / maximum starting timestep N = 1.44 (corresponding to timestep 250 in Stable Diffusion)
    Hand-selected limit on the starting timestep to enforce fidelity; directly controls the entire inversion trajectory and is not derived from first principles.
  • Loss weights lambda_l and lambda_g = lambda_l=2.0, lambda_g=0.1
    Chosen hyperparameters balancing LPIPS and GAN losses in Eq. (11); Table 6 shows they strongly affect perceptual metrics.
  • Training starting-timestep set S = {250, 200, 150, 100}
    Pre-selected starting timesteps for noise predictor training; inference schedules are subsets of these plus 50.
  • Inference step schedules = e.g. {250,200,150,100,50} for 5 steps, {150,100,50} or {250,150,50} for 3 steps, {200} for the reported InvSR-1
    Sampling schedules chosen from pre-selected timesteps; the paper reports the {200} single-step configuration in the main comparison tables, which is a tuning choice.
assumptions (5)
  • standard math DDPM forward and reverse equations (Eqs. 1-4) hold for the SD-Turbo latent diffusion model.
    The method is built on the standard DDPM marginal and reverse-step formulas; no proof is given that the latent-space model exactly satisfies them.
  • domain assumption LR and HR images differ only in high-frequency details and become indistinguishable under added Gaussian noise of appropriate scale.
    Stated in Sec. 3.1 and 3.2.1 as the rationality motivation for starting from a noisy LR state; if false, the initialization premise collapses.
  • ad hoc to paper Starting at timesteps with SNR greater than 1.44 preserves fidelity.
    Enforced design choice in Sec. 3.2.1 with no independent justification beyond the reported experiments.
  • domain assumption The pre-trained SD-Turbo denoiser is accurate enough at low noise levels that intermediate noise maps can be sampled randomly.
    Sec. 3.2.3; supported by the InvSR-Int ablation in Table 5, but not derived.
  • domain assumption The quantities y0 and x0 in Eqs. (6)-(10) may be treated as SD-Turbo latent codes, with the LR image converted through the VAE encoder.
    The text defines y0 and x0 as images, yet the backbone operates in VAE latent space; the encoding step is never written down in the equations.
invented entities (1)
  • Noise predictor network f_w independent evidence
    purpose: Produces the noise map used to construct the initial diffusion state from the LR image (Eq. 8).
    A new learned component introduced by the paper; its effectiveness is evidenced by held-out benchmark results and the released model, so it is not an unfalsifiable postulate.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Arbitrary-steps Image Super-resolution via Diffusion Inversion." pith.science (2026). https://pith.science/paper/HHALNZJJ

@misc{pith2026241209013,
  author       = {Pith},
  title        = {Pith review of: Arbitrary-steps Image Super-resolution via Diffusion Inversion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HHALNZJJ}},
  note         = {Machine review of arXiv:2412.09013}
}
read the original abstract

This study presents a new image super-resolution (SR) technique based on diffusion inversion, aiming at harnessing the rich image priors encapsulated in large pre-trained diffusion models to improve SR performance. We design a Partial noise Prediction strategy to construct an intermediate state of the diffusion model, which serves as the starting sampling point. Central to our approach is a deep noise predictor to estimate the optimal noise maps for the forward diffusion process. Once trained, this noise predictor can be used to initialize the sampling process partially along the diffusion trajectory, generating the desirable high-resolution result. Compared to existing approaches, our method offers a flexible and efficient sampling mechanism that supports an arbitrary number of sampling steps, ranging from one to five. Even with a single sampling step, our method demonstrates superior or comparable performance to recent state-of-the-art approaches. The code and model are publicly available at https://github.com/zsyOAOA/InvSR.

Figures

Figures reproduced from arXiv: 2412.09013 by the authors.

Figure 1
Figure 1. Qualitative comparisons of our proposed method to recent state-of-the-art diffusion-based approaches on two real-world ex [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Inference flow of our proposed method, wherein [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. From left to right: (a) zoomed LR image, (b) pre [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Visual results of different methods on two typical real-world examples from [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: A typical visual comparison of the proposed [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]
Figure 6
Figure 6. Figure 6: Qualitative comparisons of the proposed InvSR with different sampling steps, where the number of sampling steps is annotated in the format “InvSR-Steps”. In the first example, mainly degraded by blurriness, multi-step sampling is preferable to single-step sampling as i…
Figure 7
Figure 7. Figure 7: Visual comparisons of the proposed method with various loss configurations. (a) Zoomed LR image, (b) Baseline1 with [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]
Figure 8
Figure 8. Figure 8: Visual comparisons of various methods on three typical examples from [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 9
Figure 9. Figure 9: Visual comparisons of various methods on four real-world examples from [PITH_FULL_IMAGE:figures/full_fig_p016_9.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

  1. ICME 2025 Grand Challenge on Video Super-Resolution for Video Conferencing

    eess.IV 2025-06 conditional novelty 5.0 of 10

    Under causal low-delay video conferencing SR, single-image baselines won the general and talking-head tracks, OCR-based text recovery won the screen-content track, and PSNR/SSIM correlated weakly with subjective quality.

Reference graph

Works this paper leans on

75 extracted references · 66 canonical work pages · cited by 1 Pith paper

  1. [1]

    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

  2. [2]

    The 2018 pirm challenge on per- ceptual image super-resolution

    Yochai Blau, Roey Mechrez, Radu Timofte, Tomer Michaeli, and Lihi Zelnik-Manor. The 2018 pirm challenge on per- ceptual image super-resolution. In Eur. Conf. Comput. Vis. Worksh., pages 0–0, 2018. 6

  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. InInt. Conf. Comput. Vis., pages 3086–3095, 2019. 6

  4. [4]

    Glean: Generative latent bank for im- age super-resolution and beyond

    Kelvin CK Chan, Xiangyu Xu, Xintao Wang, Jinwei Gu, and Chen Change Loy. Glean: Generative latent bank for im- age super-resolution and beyond. IEEE Trans. Pattern Anal. Mach. Intell., 45(3):3154–3168, 2022. 2, 5

  5. [5]

    Blind image restoration via fast diffusion inversion

    Hamadi Chihaoui, Abdelhak Lemkhenter, and Paolo Favaro. Blind image restoration via fast diffusion inversion. In Adv. Neural Inform. Process. Syst., 2024. 3

  6. [6]

    Improving diffusion models for inverse prob- lems using manifold constraints

    Hyungjin Chung, Byeongsu Sim, Dohoon Ryu, and Jong Chul Ye. Improving diffusion models for inverse prob- lems using manifold constraints. In Adv. Neural Inform. Pro- cess. Syst., pages 25683–25696, 2022. 3

  7. [7]

    Come-closer-diffuse-faster: Accelerating conditional diffu- sion models for inverse problems through stochastic con- traction

    Hyungjin Chung, Byeongsu Sim, and Jong Chul Ye. Come-closer-diffuse-faster: Accelerating conditional diffu- sion models for inverse problems through stochastic con- traction. In IEEE Conf. Comput. Vis. Pattern Recog., pages 12413–12422, 2022. 2, 3

  8. [8]

    Diffusion pos- terior sampling for general noisy inverse problems

    Hyungjin Chung, Jeongsol Kim, Michael Thompson Mc- cann, Marc Louis Klasky, and Jong Chul Ye. Diffusion pos- terior sampling for general noisy inverse problems. In Int. Conf. Learn. Represent., 2023. 3

Show all 75 references
  1. [9]

    Imagenet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In IEEE Conf. Comput. Vis. Pattern Recog., pages 248–255, 2009. 6

  2. [10]

    Taming transformers for high-resolution image synthesis

    Patrick Esser, Robin Rombach, and Bjorn Ommer. Taming transformers for high-resolution image synthesis. In IEEE Conf. Comput. Vis. Pattern Recog. , pages 12873–12883,

  3. [11]

    Generative diffu- sion prior for unified image restoration and enhancement

    Ben Fei, Zhaoyang Lyu, Liang Pan, Junzhe Zhang, Weidong Yang, Tianyue Luo, Bo Zhang, and Bo Dai. Generative diffu- sion prior for unified image restoration and enhancement. In IEEE Conf. Comput. Vis. Pattern Recog., pages 9935–9946,

  4. [12]

    An image is worth one word: Personalizing text-to- image generation using textual inversion

    Rinon Gal, Yuval Alaluf, Yuval Atzmon, Or Patash- nik, Amit H Bermano, Gal Chechik, and Daniel Cohen- Or. An image is worth one word: Personalizing text-to- image generation using textual inversion. arXiv preprint arXiv:2208.01618, 2022. 3

  5. [13]

    ReNoise: Real image inversion through iterative noising

    Daniel Garibi, Or Patashnik, Andrey V oynov, Hadar Averbuch-Elor, and Daniel Cohen-Or. ReNoise: Real image inversion through iterative noising. In Eur. Conf. Comput. Vis., 2024. 2, 3

  6. [14]

    Generative adversarial nets

    Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets. Adv. Neural In- form. Process. Syst., 27, 2014. 2, 5, 12

  7. [15]

    Image processing using multi-code gan prior

    Jinjin Gu, Yujun Shen, and Bolei Zhou. Image processing using multi-code gan prior. In IEEE Conf. Comput. Vis. Pat- tern Recog., pages 3012–3021, 2020. 2

  8. [16]

    Denoising dif- fusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising dif- fusion probabilistic models. In Adv. Neural Inform. Process. Syst., pages 6840–6851, 2020. 2, 3, 5

  9. [17]

    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 IEEE Conf. Comput. Vis. Pattern Recog., pages 5197–5206,

  10. [18]

    Dslr-quality photos on mo- bile devices with deep convolutional networks

    Andrey Ignatov, Nikolay Kobyshev, Radu Timofte, Kenneth Vanhoey, and Luc Van Gool. Dslr-quality photos on mo- bile devices with deep convolutional networks. In Int. Conf. Comput. Vis., 2017. 6

  11. [19]

    Pnp inversion: Boosting diffusion-based editing with 3 lines of code

    Xuan Ju, Ailing Zeng, Yuxuan Bian, Shaoteng Liu, and Qiang Xu. Pnp inversion: Boosting diffusion-based editing with 3 lines of code. In Int. Conf. Learn. Represent., 2024. 3

  12. [20]

    Eta inver- sion: Designing an optimal eta function for diffusion-based real image editing

    Wonjun Kang, Kevin Galim, and Hyung Il Koo. Eta inver- sion: Designing an optimal eta function for diffusion-based real image editing. In Eur. Conf. Comput. Vis., 2024. 3

  13. [21]

    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. In IEEE Conf. Comput. Vis. Pattern Recog., pages 4401–4410,

  14. [22]

    Elucidating the design space of diffusion-based generative models

    Tero Karras, Miika Aittala, Timo Aila, and Samuli Laine. Elucidating the design space of diffusion-based generative models. In Adv. Neural Inform. Process. Syst., pages 26565– 26577, 2022. 2, 4

  15. [23]

    Denoising diffusion restoration models

    Bahjat Kawar, Michael Elad, Stefano Ermon, and Jiaming Song. Denoising diffusion restoration models. In Adv. Neu- ral Inform. Process. Syst., pages 23593–23606, 2022. 2, 3

  16. [24]

    Musiq: Multi-scale image quality transformer

    Junjie Ke, Qifei Wang, Yilin Wang, Peyman Milanfar, and Feng Yang. Musiq: Multi-scale image quality transformer. In Int. Conf. Comput. Vis., pages 5148–5157, 2021. 6

  17. [25]

    Auto-encoding variational bayes

    Diederik P Kingma. Auto-encoding variational bayes. In Int. Conf. Learn. Represent., 2014. 3

  18. [26]

    Kingma and Jimmy Ba

    Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. In Int. Conf. Learn. Represent. ,

  19. [27]

    Distillation-free one-step diffusion for real-world image super-resolution

    Jianze Li, Jiezhang Cao, Zichen Zou, Xiongfei Su, Xin Yuan, Yulun Zhang, Yong Guo, and Xiaokang Yang. Distillation-free one-step diffusion for real-world image super-resolution. arXiv preprint arXiv:2410.04224, 2024. 3

  20. [28]

    Ls- dir: 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, Rakesh Ranjan, Radu Timofte, and Luc Van Gool. Ls- dir: A large scale dataset for image restoration. In IEEE Conf. Comput. Vis. Pattern Recog. Worksh. , pages...

  21. [29]

    Common diffusion noise schedules and sample steps are flawed

    Shanchuan Lin, Bingchen Liu, Jiashi Li, and Xiao Yang. Common diffusion noise schedules and sample steps are flawed. In Proceedings of the IEEE/CVF Winter Conf. on Applications of Comput. Vision, pages 5404–5411, 2024. 4

  22. [30]

    DiffBIR: Towards blind image restoration with generative diffusion prior

    Xinqi Lin, Jingwen He, Ziyan Chen, Zhaoyang Lyu, Ben Fei, Bo Dai, Wanli Ouyang, Yu Qiao, and Chao Dong. DiffBIR: Towards blind image restoration with generative diffusion prior. arXiv preprint arXiv:2308.15070, 2023. 2, 3, 6, 7

  23. [31]

    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 Int. Conf. Comput. Vis. , pages 416–423, 2001

  24. [32]

    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. Mul- timedia Tools and Applications, 76:21811–21838, 2017. 6

  25. [33]

    Fixed-point inversion for text-to- image diffusion models

    Barak Meiri, Dvir Samuel, Nir Darshan, Gal Chechik, Shai Avidan, and Rami Ben-Ari. Fixed-point inversion for text-to- image diffusion models. arXiv preprint arXiv:2312.12540,

  26. [34]

    completely blind

    Anish Mittal, Rajiv Soundararajan, and Alan C Bovik. Mak- ing a “completely blind” image quality analyzer. IEEE Sig- nal processing letters, 20(3):209–212, 2012. 6

  27. [35]

    Negative-prompt inversion: Fast image inversion for editing with text-guided diffusion models

    Daiki Miyake, Akihiro Iohara, Yu Saito, and Toshiyuki Tanaka. Negative-prompt inversion: Fast image inversion for editing with text-guided diffusion models. arXiv preprint arXiv:2305.16807, 2023. 3

  28. [36]

    Null-text inversion for editing real images using guided diffusion models

    Ron Mokady, Amir Hertz, Kfir Aberman, Yael Pritch, and Daniel Cohen-Or. Null-text inversion for editing real images using guided diffusion models. In IEEE Conf. Comput. Vis. Pattern Recog., pages 6038–6047, 2023. 2, 3

  29. [37]

    Metric learning based interactive modulation for real-world super-resolution

    Chong Mou, Yanze Wu, Xintao Wang, Chao Dong, Jian Zhang, and Ying Shan. Metric learning based interactive modulation for real-world super-resolution. In Eur. Conf. Comput. Vis., pages 723–740. Springer, 2022. 12

  30. [38]

    Gibbsddrm: A partially collapsed gibbs sampler for solving blind inverse problems with denoising diffusion restoration

    Naoki Murata, Koichi Saito, Chieh-Hsin Lai, Yuhta Takida, Toshimitsu Uesaka, Yuki Mitsufuji, and Stefano Ermon. Gibbsddrm: A partially collapsed gibbs sampler for solving blind inverse problems with denoising diffusion restoration. In Int. Conf. Mach. Learn. , pages 25501–25522. PMLR,

  31. [39]

    Visual instruction inversion: Image editing via image prompting

    Thao Nguyen, Yuheng Li, Utkarsh Ojha, and Yong Jae Lee. Visual instruction inversion: Image editing via image prompting. In Adv. Neural Inform. Process. Syst., 2024. 3

  32. [40]

    You only need one step: Fast super-resolution with stable diffusion via scale distillation

    Mehdi Noroozi, Isma Hadji, Brais Martinez, Adrian Bulat, and Georgios Tzimiropoulos. You only need one step: Fast super-resolution with stable diffusion via scale distillation. In Eur. Conf. Comput. Vis., 2024. 3

  33. [41]

    Exploiting deep genera- tive prior for versatile image restoration and manipulation

    Xingang Pan, Xiaohang Zhan, Bo Dai, Dahua Lin, Chen Change Loy, and Ping Luo. Exploiting deep genera- tive prior for versatile image restoration and manipulation. IEEE Trans. Pattern Anal. Mach. Intell., 44(11):7474–7489,

  34. [42]

    Pytorch: An imperative style, high-performance deep learning library

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al. Pytorch: An imperative style, high-performance deep learning library. In Adv. Neural Inform. Process. Syst., 2019. 6

  35. [43]

    High-resolution image syn- thesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj¨orn Ommer. High-resolution image syn- thesis with latent diffusion models. In IEEE Conf. Comput. Vis. Pattern Recog., pages 10684–10695, 2022. 2, 4, 6, 7, 12

  36. [44]

    Adversarial diffusion distillation

    Axel Sauer, Dominik Lorenz, Andreas Blattmann, and Robin Rombach. Adversarial diffusion distillation. In Eur. Conf. Comput. Vis., pages 87–103. Springer, 2024. 5

  37. [45]

    Deep unsupervised learning using nonequilibrium thermodynamics

    Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli. Deep unsupervised learning using nonequilibrium thermodynamics. In Int. Conf. Mach. Learn., pages 2256–2265. PMLR, 2015. 2, 3

  38. [46]

    Denois- ing diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denois- ing diffusion implicit models. In Int. Conf. Learn. Repre- sent., 2021. 2, 3, 4

  39. [47]

    Pseudoinverse-guided diffusion models for inverse problems

    Jiaming Song, Arash Vahdat, Morteza Mardani, and Jan Kautz. Pseudoinverse-guided diffusion models for inverse problems. In Int. Conf. Learn. Represent., 2023. 3

  40. [48]

    Score-based generative modeling through stochastic differential equa- tions

    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. In Int. Conf. Learn. Represent., 2021. 2, 3

  41. [49]

    Coser: Bridging image and language for cognitive super-resolution

    Haoze Sun, Wenbo Li, Jianzhuang Liu, Haoyu Chen, Ren- jing Pei, Xueyi Zou, Youliang Yan, and Yujiu Yang. Coser: Bridging image and language for cognitive super-resolution. In IEEE Conf. Comput. Vis. Pattern Recog. , pages 25868– 25878, 2024. 3

  42. [50]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszko- reit, Llion Jones, Aidan N Gomez, Ł ukasz Kaiser, and Illia Polosukhin. Attention is all you need. InAdv. Neural Inform. Process. Syst., 2017. 6

  43. [51]

    Edict: Ex- act diffusion inversion via coupled transformations

    Bram Wallace, Akash Gokul, and Nikhil Naik. Edict: Ex- act diffusion inversion via coupled transformations. In IEEE Conf. Comput. Vis. Pattern Recog. , pages 22532–22541,

  44. [52]

    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. In AAAI, 2023. 6

  45. [53]

    Chan, and Chen Change Loy

    Jianyi Wang, Zongsheng Yue, Shangchen Zhou, Kelvin C.K. Chan, and Chen Change Loy. Exploiting diffusion prior for real-world image super-resolution. Int. J. Comput. Vis., pages 1–21, 2024. 3, 6, 7

  46. [54]

    Recovering realistic texture in image super-resolution by deep spatial feature transform

    Xintao Wang, Ke Yu, Chao Dong, and Chen Change Loy. Recovering realistic texture in image super-resolution by deep spatial feature transform. In IEEE Conf. Comput. Vis. Pattern Recog., 2018. 3

  47. [55]

    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. InInt. Conf. Comput. Vis. Worksh., pages 1905–1914, 2021. 5, 6, 7

  48. [56]

    Zero-shot image restoration using denoising diffusion null-space model

    Yinhuai Wang, Jiwen Yu, and Jian Zhang. Zero-shot image restoration using denoising diffusion null-space model. In Int. Conf. Learn. Represent., 2023. 2, 3

  49. [57]

    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. In IEEE Conf. Comput. Vis. Pat- tern Recog., pages 25796–25805, 2024. 6, 7

  50. [58]

    Bovik, H.R

    Zhou Wang, A.C. Bovik, H.R. Sheikh, and E.P. Simoncelli. Image quality assessment: from error visibility to struc- tural similarity. IEEE Trans. Image Process., 13(4):600–612,

  51. [59]

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

    Rongyuan Wu, Lingchen Sun, Zhiyuan Ma, and Lei Zhang. One-step effective diffusion network for real-world image super-resolution. In Adv. Neural Inform. Process. Syst. ,

  52. [60]

    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. In IEEE Conf. Comput. Vis. Pattern Recog., pages 25456–25467, 2024. 2, 3, 5, 6, 7

  53. [61]

    Diffir: Efficient diffusion model for image restoration

    Bin Xia, Yulun Zhang, Shiyin Wang, Yitong Wang, Xing- long Wu, Yapeng Tian, Wenming Yang, and Luc Van Gool. Diffir: Efficient diffusion model for image restoration. InInt. Conf. Comput. Vis., pages 13095–13105, 2023. 3

  54. [62]

    Gan inversion: A survey

    Weihao Xia, Yulun Zhang, Yujiu Yang, Jing-Hao Xue, Bolei Zhou, and Ming-Hsuan Yang. Gan inversion: A survey. IEEE Trans. Pattern Anal. Mach. Intell. , 45(3):3121–3138,

  55. [63]

    DreamClean: Restoring clean image using deep diffusion prior

    Jie Xiao, Ruili Feng, Han Zhang, Zhiheng Liu, Zhantao Yang, Yurui Zhu, Xueyang Fu, Kai Zhu, Yu Liu, and Zheng- Jun Zha. DreamClean: Restoring clean image using deep diffusion prior. In Int. Conf. Learn. Represent., 2024. 3

  56. [64]

    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. In Eur. Conf. Comput. Vis., 2024. 2, 3

  57. [65]

    Im- proved distribution matching distillation for fast image syn- thesis

    Tianwei Yin, Micha ¨el Gharbi, Taesung Park, Richard Zhang, Eli Shechtman, Fredo Durand, and William T Freeman. Im- proved distribution matching distillation for fast image syn- thesis. In Adv. Neural Inform. Process. Syst., 2024. 5

  58. [66]

    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 IEEE Conf. Com- put. Vis. Pattern Recog., pages 25669–25680, 2024. 3

  59. [67]

    DifFace: Blind face restoration with diffused error contraction

    Zongsheng Yue and Chen Change Loy. DifFace: Blind face restoration with diffused error contraction. IEEE Trans. Pat- tern Anal. Mach. Intell., 2024. 3

  60. [68]

    Resshift: Efficient diffusion model for image super- resolution by residual shifting

    Zongsheng Yue, Jianyi Wang, and Chen Change Loy. Resshift: Efficient diffusion model for image super- resolution by residual shifting. In Adv. Neural Inform. Pro- cess. Syst., pages 13294–13307, 2023. 6

  61. [69]

    Effi- cient diffusion model for image restoration by residual shift- ing

    Zongsheng Yue, Jianyi Wang, and Chen Change Loy. Effi- cient diffusion model for image restoration by residual shift- ing. IEEE Trans. Pattern Anal. Mach. Intell., 2024. 6, 7

  62. [70]

    On single image scale-up using sparse-representations

    Roman Zeyde, Michael Elad, and Matan Protter. On single image scale-up using sparse-representations. In Int. Conf. on Curves and Surfaces, pages 711–730. Springer, 2012. 6

  63. [71]

    Degradation-guided one-step im- age super-resolution with diffusion priors

    Aiping Zhang, Zongsheng Yue, Renjing Pei, Wenqi Ren, and Xiaochun Cao. Degradation-guided one-step im- age super-resolution with diffusion priors. arXiv preprint arXiv:2409.17058, 2024. 3

  64. [72]

    Exact diffusion inversion via bidirectional integration ap- proximation

    Guoqiang Zhang, Jonathan P Lewis, and W Bastiaan Kleijn. Exact diffusion inversion via bidirectional integration ap- proximation. In Eur. Conf. Comput. Vis., pages 19–36, 2024. 3

  65. [73]

    Designing a practical degradation model for deep blind image super-resolution

    Kai Zhang, Jingyun Liang, Luc Van Gool, and Radu Timo- fte. Designing a practical degradation model for deep blind image super-resolution. In Int. Conf. Comput. Vis. , pages 4791–4800, 2021. 5, 6, 7

  66. [74]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shechtman, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. In IEEE Conf. Comput. Vis. Pattern Recog., pages 586–595, 2018. 5, 6, 12

  67. [75]

    InvSR-Int

    Jiapeng Zhu, Yujun Shen, Deli Zhao, and Bolei Zhou. In- domain gan inversion for real image editing. In Eur. Conf. Comput. Vis., pages 592–608. Springer, 2020. 2 Arbitrary-steps Image Super-resolution via Diffusion Inversion Supplementary Material This supplemental material ma...

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.