Pith. sign in

REVIEW 4 major objections 5 minor 3 cited by

ClusIR claims that making an all-in-one restorer explicitly recognize degradation clusters before choosing experts yields state-of-the-art unified restoration, with 33.06 dB average PSNR on noise, haze, and rain.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-03 16:57 UTC pith:PNJNZU6P

load-bearing objection Solid all-in-one restoration paper with real but incremental gains; the central claim about PCGRM is undermined by the paper's own One-by-One results. the 4 major comments →

arxiv 2512.10948 v2 pith:PNJNZU6P submitted 2025-12-11 cs.CV

ClusIR: Towards Cluster-Guided All-in-One Image Restoration

Shengkai Hu , Jiaqi Ma , Xu Zhang , Yongcheng Jing , Lefei Zhang , Jun Wan This is my paper
classification cs.CV
keywords All-in-one image restorationmixture of expertsdegradation-aware routingcluster prototypesfrequency modulationimage denoisingimage dehazingimage deraining
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

ClusIR argues that all-in-one image restoration improves when the network first identifies which kind of degradation it faces, then chooses specialized experts conditioned on that identification. The paper's route is a two-stage probabilistic router: a softmax over learned degradation clusters followed by a cluster-conditional softmax over experts, paired with a frequency module that uses cluster-derived cues to adjust low-frequency structure and high-frequency texture separately. On the three-task benchmark (denoising, dehazing, deraining) the system reports 33.06 dB average PSNR and 0.923 SSIM, ahead of the compared unified methods, and on the five-task setting it matches the best PSNR while leading SSIM. A sympathetic reader would take the contribution as evidence that explicit degradation semantics, not just implicit expert competition, is a useful organizing principle for unified restoration.

Core claim

The paper claims that factorizing the router into two stages — first a degradation-aware cluster posterior p(C=c|x), then a cluster-conditional expert posterior p(E=e|c,x) — separates degradation recognition from expert activation. Flat single-softmax routers, it argues, force a unimodal decision space that entangles mixed degradations and creates unstable global expert competition. The two-stage form yields a multimodal distribution over experts and, combined with cluster-guided frequency modulation, gives the best reported average performance among compared all-in-one methods on the three-task benchmark (33.06 dB / 0.923) and strong composite-degradation results on CDD11 (27.33 dB / 0.878)

What carries the argument

Probabilistic Cluster-Guided Routing Mechanism (PCGRM): a two-stage router that first computes a softmax posterior over learned degradation clusters (spherically normalized, nearly orthogonal prototypes per encoder stage), then computes a cluster-conditional softmax over experts inside the top-K clusters, so expert choice is conditioned on a recognized degradation type. The Degradation-Aware Frequency Modulation Module (DAFMM) takes the cluster-derived prompt and applies a learnable low-pass filter (Frequency Self-Mining Block) plus Fourier amplitude-phase fusion to refine low frequencies and channel-gated modulation for high frequencies. Together they carry the claim that semantic degradati

Load-bearing premise

The load-bearing premise is that two-stage cluster-then-expert routing genuinely disentangles degradation recognition from expert activation rather than re-parameterizing a single softmax; the paper's own Sec. 4.4 concedes that in one-by-one training the cluster prototypes fail to activate yet results remain strong, so the causal role of the clusters is not yet established.

What would settle it

Train the identical architecture but replace PCGRM with a single flat softmax router over the same experts, holding parameter count roughly equal, and compare on CDD11 mixed degradations; if the flat router matches or beats ClusIR, the two-stage factorization is not doing the work. Alternatively, feed images with known single degradations and inspect the cluster posterior: if one cluster dominates regardless of degradation type, the clusters are not encoding degradation semantics.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Unified restoration networks no longer need to trade away per-degradation specialization: the same weights can route differently for noise, haze, and rain while sharing most of the backbone.
  • Cluster assignments give a built-in, testable description of what degradation the model thinks it is seeing, which can be read out at inference time.
  • Because routing is conditioned on clusters, adding a new degradation type or benchmark may require only a new prototype rather than a retrained router.
  • Frequency modulation tied to cluster cues lets the same model emphasize structural low frequencies for blur-like degradations and texture high frequencies for noise-like ones.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the clusters genuinely correspond to degradation types, the mechanism doubles as a degradation classifier; one could use the cluster posterior to report "this image looks hazy" with no extra head. The paper does not exploit this, but it follows directly from the design.
  • The two-stage factorization is an instance of a general cost-geometry idea: first partition the input space coarsely, then act within the partition. Similar cluster-conditioned routing could apply to other inverse problems such as super-resolution or compression-artifact removal, where mixed corruptions are common.
  • The one-by-one result the paper reports — strong performance even when clusters do not activate — suggests the gains on single degradations may come substantially from added capacity, and the cluster mechanism's distinctive value is probably confined to mixed and composite degradations.
  • A controlled test that equalizes parameter count between the two-stage router and a single softmax router over the same expert set would isolate whether the factorization itself, rather than extra parameters, explains the benchmark gaps.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. ClusIR proposes a cluster-guided all-in-one image restoration framework. Its two main components are a Probabilistic Cluster-Guided Routing Mechanism (PCGRM), which factorizes expert routing into a degradation-cluster posterior p(C=c|x) and a cluster-conditional expert posterior p(E=e|c,x), and a Degradation-Aware Frequency Modulation Module (DAFMM), which uses cluster prompts with DWT and a learnable Frequency Self-Mining Block for frequency decomposition. The paper reports experiments on three-degradation (N+H+R) and five-degradation (N+H+R+B+L) settings, on the composite CDD11 benchmark in the supplementary material, and on One-by-One single-task training. Table 1 reports an average PSNR/SSIM of 33.06/0.923 for the N+H+R setting, claimed to surpass recent all-in-one methods; Table 2 reports 30.58/0.919 on five tasks. The central claim is that PCGRM's hierarchical two-stage routing disentangles degradation recognition from expert activation and thereby delivers state-of-the-art adaptive restoration.

Significance. If the mechanism's causal role were established, ClusIR would offer a useful design pattern for AiOIR: explicit learnable cluster prototypes with two-stage routing, coupled with frequency-domain modulation. The empirical strengths are the external benchmark evaluations (Tables 1–3 and supplementary CDD11 results), the component ablations, the orthogonal-initialization study, and the efficiency variants (ClusIR-T/S/Base). The gains over strong recent methods such as MoCE-IR, DFPIR, and AdaIR are modest but consistent, especially in average PSNR/SSIM. However, the load-bearing attribution of these gains to PCGRM is not yet supported: the paper's own Sec. 4.4 states that cluster prototypes fail to activate in One-by-One training, and the main ablation in Table 4 adds PCGRM-MoE as a whole package without isolating the two-stage routing from added parameters. The result is promising but needs substantial verification before the 'state-of-the-art via cluster-guided routing' claim is accepted.

major comments (4)
  1. [§4.4 and Table 3] The paper's own Sec. 4.4 states that in One-by-One training ClusIR 'fails to effectively activate the cluster prototypes, which undermines the PCGRM, prevents reliable degradation discrimination, and consequently limits the generation of semantic prompts.' Table 3 nonetheless reports best AiOIR results for Kodak24 (35.06 dB at sigma=15), SOTS (32.67/0.983), and LOL (23.82/0.852). This is an internal inconsistency that is load-bearing: Contribution (1)/(2) attributes state-of-the-art performance to the cluster-guided design, but the single-task SOTA cannot be attributed to PCGRM if the mechanism is inert. Please add controlled results under One-by-One with PCGRM ablated (WTB+DAFMM only) and with a matched-capacity single-stage MoE, and quantify cluster activation (e.g., entropy or purity against task labels) rather than relying only on t-SNE.
  2. [§3.2, Eqs. (1)–(2)] The central motivation that a flat softmax router is 'unimodal' and therefore collapses mixed degradations is not established. A softmax over experts is a categorical distribution over the expert set; for different inputs x it can place high mass on different experts, so the router as a function of x can represent multimodal degradation mixtures. The limitation is a capacity/expressivity matter, not a consequence of unimodality. Please provide a formal argument or a controlled experiment comparing the two-stage PCGRM with a single-stage softmax router of matched parameters; otherwise the stated advantage of the factorization is a rhetorical contrast rather than a verified property.
  3. [Table 4 (ablation)] The ablation from (a) to (b) adds PCGRM-MoE, which includes all expert parameters and gating weights, yielding +0.27 dB. This does not isolate the two-stage probabilistic routing from the increase in model capacity. A single-stage MoE baseline (Eq. (1)) with the same number of experts and parameters would disentangle capacity from the routing structure. Without such a baseline, the result is consistent with the hypothesis that the gain comes from extra parameters rather than from cluster-guided disentanglement.
  4. [§4.3, Figs. 5–6] The claim that clusters encode degradation semantics rests on t-SNE and affinity maps generated by the same model. These visualizations show separation but not that the clusters correspond to the degradation classes used in training/testing, which would be necessary for 'explicit degradation semantics.' Please report quantitative alignment measures (e.g., cluster purity or NMI with task labels, or confusion between noise levels and tasks) and, ideally, a probe of p(C|x) against held-out degradation labels.
minor comments (5)
  1. [Table 2] The row labeled 'InstructIR [28] ECCV'24' cites reference [28] (Up-Restorer); InstructIR is reference [8]. Please correct the citation.
  2. [§3.2, Eq. (7)] Eq. (7) sums the Gaussian prompt over all N prototypes, while Eqs. (5)–(6) normalize over the top-K1 selected clusters. Please make the summation range consistent (selected set vs. full bank) and define sigma_c and epsilon_c precisely.
  3. [§6.1 / Table 3] The supplementary says denoising is evaluated on CBSD68 and Urban100, while Table 3 of the main text reports Kodak24. Please state which benchmark is used for One-by-One denoising and include Urban100 results or remove the mention.
  4. [Figs. 6 and 9] The figure captions 'Visual degradation of PCGRM-MoE' likely mean 'Visualization of PCGRM-MoE activations/affinity'; please reword.
  5. [§3.3] Eq. (14)'s IFFT(,) notation and Eq. (15)'s channel-wise gating are not fully defined; specify dimensions and the activation function used.

Circularity Check

0 steps flagged

No significant circularity; the central claims rest on external benchmarks and not on fitted parameters or self-citations.

full rationale

Walking the derivation chain, no load-bearing step reduces to its own inputs. The PCGRM factorization in Eqs. (3)-(10) is a strictly more expressive mixture over cluster-conditional expert softmaxes; it is not equivalent by construction to the flat softmax router in Eq. (1). The reported SOTA numbers (e.g., 33.06 dB/0.923 in Table 1) are measured against external benchmarks (CBSD68, SOTS, Rain100L, GoPro, LOL), so no fitted parameter is renamed as a prediction. The ablation in Table 4 attributes +0.27 dB to PCGRM-MoE, but that is a standard component ablation rather than a circular reduction. The paper's own Sec. 4.4 says that under One-by-One training ClusIR 'fails to effectively activate the cluster prototypes' while Table 3 reports strong single-task results; this is an internal mechanistic/attribution inconsistency and a correctness risk, not circularity, because no equation or fitted constant forces the outcome. The self-citations (Perceive-IR [72], ProRes [31]) appear as comparison baselines, with Table 1 noting 'Partial results are referenced from Perceive-IR [72]'; they are not used to justify ClusIR's mechanism or to prohibit alternatives, so they are not load-bearing. The t-SNE and affinity-map analyses are produced by the same trained model and therefore give weak independent support for the semantic-cluster interpretation, but weak evidence is not circular derivation. Overall, the paper's derivations are self-contained and its headline claims are externally checkable.

Axiom & Free-Parameter Ledger

6 free parameters · 6 axioms · 3 invented entities

The central claim rests on hand-chosen cluster/active-expert counts, an assumed correspondence between prototypes and degradation semantics, benchmark validity, and unspecified training details. No code, weights, or error bars are provided, so independent verification is not yet possible.

free parameters (6)
  • Cluster prototype counts per stage = [3, 3, 3, 3]
    Hand-chosen; Tab. 5 ablation shows unbalanced configurations hurt performance, so this choice is load-bearing.
  • Active cluster count K1 per stage = [2, 2, 2, 2]
    Hand-chosen top-K; no sensitivity analysis beyond cluster counts.
  • Active expert count K2 per cluster = Not stated
    Eq. 10 depends on K2, but the paper never reports the number of experts or how many are active.
  • Data expansion ratios for denoising/deraining/deblurring/low-light = 3, 120, 5, 200
    Adopted from AdaIR [11]; controls class balance and affects average results.
  • L1/MS-SSIM loss weights = Unspecified (equal)
    Sec. 4.1 says joint objective but no weighting; affects optimization and final PSNR/SSIM.
  • Orthogonal vs random prototype initialization = Orthogonal
    Tab. 9 shows orthogonal initialization outperforms random by 0.22 dB; a design choice selected by validation.
axioms (6)
  • ad hoc to paper A softmax over expert indices is a unimodal distribution that collapses mixed degradations into a single mode.
    Sec. 3.2 uses this to motivate PCGRM (Eqs. 1-3). A categorical softmax is not a unimodal distribution over an ordered/continuous space; this is a conceptual assumption, not derived.
  • domain assumption Learned cluster prototypes capture latent degradation semantics from global average pooled features.
    Sec. 3.2 Stage 1 and Figs. 5-6 assume cosine similarity to prototype banks produces degradation-type clusters; no external label or benchmark verifies cluster meaning.
  • domain assumption DWT subbands retain residual mid-frequency structure that a learnable low-pass filter (FSB) can separate.
    Sec. 3.3 and Fig. 3 motivate FSB; this is an empirical modeling choice, not proven by the framework.
  • domain assumption PSNR/SSIM on CBSD68, SOTS, Rain100L, GoPro, LOL, and CDD11 are valid measures of restoration quality and sufficient to compare methods.
    Sec. 4 uses these benchmarks as ground truth for all claims; standard in the field but not proven.
  • domain assumption Wavelet-based Transformer Block (WTB) from Wave-ViT is a strong base restoration backbone, and adding PCGRM/DAFMM should improve it.
    Sec. 3.1 builds ClusIR on WTB and PGB; baseline validity is assumed from [59].
  • domain assumption Training with L1 + MS-SSIM for 150 epochs at batch size 48 is sufficient for convergence across all tasks.
    Sec. 4.1 states this schedule; no convergence analysis is provided.
invented entities (3)
  • Learnable cluster prototype banks P^(l) no independent evidence
    purpose: Represent layer-specific degradation semantics for routing
    No external/falsifiable signal confirms these prototypes correspond to degradations; evidence is internal similarity/t-SNE analysis (Sec. 4.3).
  • Gaussian semantic prior N(mu_c, diag sigma_c^2) per cluster no independent evidence
    purpose: Model intra-cluster uncertainty and generate degradation prompts via reparameterization
    Introduced in Eq. 7; no testable prediction outside the network.
  • Frequency Self-Mining Block (FSB) no independent evidence
    purpose: Learn data-driven low-pass filter to separate low/high frequency components
    Architectural component; its benefit is only demonstrated through end-task performance.

pith-pipeline@v1.3.0-alltime-deepseek · 19659 in / 15284 out tokens · 131312 ms · 2026-08-03T16:57:42.951395+00:00 · methodology

0 comments
read the original abstract

All-in-One Image Restoration (AiOIR) aims to recover high-quality images from diverse degradations within a unified framework. However, existing methods often fail to explicitly model degradation types and struggle to adapt their restoration behavior to complex or mixed degradations. To address these issues, we propose ClusIR, a Cluster-Guided Image Restoration framework that explicitly models degradation semantics through learnable clustering and propagates cluster-aware cues across spatial and frequency domains for adaptive restoration. Specifically, ClusIR comprises two key components: a Probabilistic Cluster-Guided Routing Mechanism (PCGRM) and a Degradation-Aware Frequency Modulation Module (DAFMM). The proposed PCGRM disentangles degradation recognition from expert activation, enabling discriminative degradation perception and stable expert routing. Meanwhile, DAFMM leverages the cluster-guided priors to perform adaptive frequency decomposition and targeted modulation, collaboratively refining structural and textural representations for higher restoration fidelity. The cluster-guided synergy seamlessly bridges semantic cues with frequency-domain modulation, empowering ClusIR to attain remarkable restoration results across a wide range of degradations. Extensive experiments on diverse benchmarks validate that ClusIR reaches competitive performance under several scenarios.

Figures

Figures reproduced from arXiv: 2512.10948 by Jiaqi Ma, Jun Wan, Lefei Zhang, Shengkai Hu, Xu Zhang, Yongcheng Jing.

Figure 1
Figure 1. Figure 1: Comparison of common image restoration paradigms. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the proposed ClusIR framework, which integrates the Probabilistic Cluster-Guided Routing Mechanism (PCGRM) [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Motivation of the Frequency Self-Mining Block (FSB). [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Visual comparison of ClusIR with SOTA AiOIR methods under the “ [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Visual degradation of PCGRM-MoE under the [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: t-SNE visualization of cross-stage feature distributions [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Pairwise prototype MSE across stages under orthogonal [PITH_FULL_IMAGE:figures/full_fig_p010_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Visual degradation of PCGRM-MoE under the ( [PITH_FULL_IMAGE:figures/full_fig_p012_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Denoising (σ = 15) visual comparisons of ClusIR with state-of-the-art All-in-One methods under One-by-One setting [PITH_FULL_IMAGE:figures/full_fig_p013_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Denoising (σ = 25) visual comparisons of ClusIR with state-of-the-art All-in-One methods under One-by-One setting [PITH_FULL_IMAGE:figures/full_fig_p013_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Denoising (σ = 50) visual comparisons of ClusIR with state-of-the-art All-in-One methods under One-by-One setting. 13 [PITH_FULL_IMAGE:figures/full_fig_p013_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Deraining visual comparisons of ClusIR with state-of-the-art All-in-One methods under One-by-One setting. [PITH_FULL_IMAGE:figures/full_fig_p014_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Dehazing visual comparisons of ClusIR with state-of-the-art All-in-One methods under One-by-One setting. [PITH_FULL_IMAGE:figures/full_fig_p014_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Denoising (σ = 15) visual comparisons of ClusIR with state-of-the-art All-in-One methods under “N+H+R” setting. 14 [PITH_FULL_IMAGE:figures/full_fig_p014_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Denoising (σ = 25) visual comparisons of ClusIR with state-of-the-art All-in-One methods under “N+H+R” setting [PITH_FULL_IMAGE:figures/full_fig_p015_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: Denoising (σ = 50) visual comparisons of ClusIR with state-of-the-art All-in-One methods under “N+H+R” setting [PITH_FULL_IMAGE:figures/full_fig_p015_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: Deraining visual comparisons of ClusIR with state-of-the-art All-in-One methods under “ [PITH_FULL_IMAGE:figures/full_fig_p015_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: Dehazing visual comparisons of ClusIR with state-of-the-art All-in-One methods under “ [PITH_FULL_IMAGE:figures/full_fig_p016_19.png] view at source ↗
Figure 20
Figure 20. Figure 20: Denoising (σ = 25) visual comparisons of ClusIR with state-of-the-art All-in-One methods under “N+H+R+B+L” setting [PITH_FULL_IMAGE:figures/full_fig_p016_20.png] view at source ↗
Figure 21
Figure 21. Figure 21: Deraining visual comparisons of ClusIR with state-of-the-art All-in-One methods under “ [PITH_FULL_IMAGE:figures/full_fig_p016_21.png] view at source ↗
Figure 22
Figure 22. Figure 22: Dehazing visual comparisons of ClusIR with state-of-the-art All-in-One methods under “ [PITH_FULL_IMAGE:figures/full_fig_p017_22.png] view at source ↗
Figure 23
Figure 23. Figure 23: Enhancement visual comparisons of ClusIR with state-of-the-art All-in-One methods under “ [PITH_FULL_IMAGE:figures/full_fig_p017_23.png] view at source ↗
Figure 24
Figure 24. Figure 24: Deblurring visual comparisons of ClusIR with state-of-the-art All-in-One methods under “ [PITH_FULL_IMAGE:figures/full_fig_p017_24.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 3 Pith papers

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

  1. SpikeRestormer: Towards Energy-Efficient All-in-One Image Restoration via Unified Event Reasoning

    cs.CV 2026-08 conditional novelty 6.0

    A spiking neural network with subtractive and additive attention performs all-in-one image restoration in one time step, matching older ANN baselines with much lower estimated energy.

  2. CoRE-UIR: Prior-guided common and residual experts for efficient all-in-one remote sensing image restoration

    cs.CV 2026-07 conditional novelty 6.0

    CoRE-UIR achieves state-of-the-art all-in-one remote sensing image restoration with a common dense expert plus low-rank routed residual experts, improving PSNR by 1.05 dB over BaryIR at 11.83x lower latency.

  3. QuReC: All-in-One Image Restoration with Query-Specific Guidance and Local-Global Response Calibration

    cs.CV 2026-07 conditional novelty 6.0

    Per-pixel degradation prototypes plus calibrated local-global attention improve all-in-one restoration accuracy on three benchmark suites.

Reference graph

Works this paper leans on

75 extracted references · 10 linked inside Pith · cited by 3 Pith papers

  1. [1]

    Contour detection and hierarchical image seg- mentation.IEEE Trans

    Pablo Arbelaez, Michael Maire, Charless Fowlkes, and Ji- tendra Malik. Contour detection and hierarchical image seg- mentation.IEEE Trans. Pattern Anal. Mach. Intell., 33(5): 898–916, 2010. 6, 10

    2010

  2. [2]

    Dehazenet: An end-to-end system for single image haze removal.IEEE Trans

    Bolun Cai, Xiangmin Xu, Kui Jia, Chunmei Qing, and Dacheng Tao. Dehazenet: An end-to-end system for single image haze removal.IEEE Trans. Image Process., 25(11): 5187–5198, 2016. 1, 2, 7

    2016

  3. [3]

    Retinexformer: One-stage retinex- based transformer for low-light image enhancement

    Yuanhao Cai, Hao Bian, Jing Lin, Haoqian Wang, Radu Tim- ofte, and Yulun Zhang. Retinexformer: One-stage retinex- based transformer for low-light image enhancement. InProc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), pages 12504– 12513, 2023. 1, 2, 7

    2023

  4. [4]

    Emerg- ing properties in self-supervised vision transformers

    Mathilde Caron, Hugo Touvron, Ishan Misra, Herv ´e J´egou, Julien Mairal, Piotr Bojanowski, and Armand Joulin. Emerg- ing properties in self-supervised vision transformers. In Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), pages 9650–9660, 2021. 2

    2021

  5. [5]

    Simple baselines for image restoration

    Liangyu Chen, Xiaojie Chu, Xiangyu Zhang, and Jian Sun. Simple baselines for image restoration. InProc. Eur. Conf. Comput. Vis. (ECCV), pages 17–33, 2022. 1, 7

    2022

  6. [6]

    Learn- ing a sparse transformer network for effective image derain- ing

    Xiang Chen, Hao Li, Mingqiang Li, and Jinshan Pan. Learn- ing a sparse transformer network for effective image derain- ing. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recog- nit. (CVPR), pages 5896–5905, 2023. 7

    2023

  7. [7]

    Hierarchical integration diffu- sion model for realistic image deblurring

    Zheng Chen, Yulun Zhang, Ding Liu, Bin Xia, Jinjin Gu, Linghe Kong, and Xin Yuan. Hierarchical integration diffu- sion model for realistic image deblurring. InAdv. Neural Inf. Process. Syst. (NeurIPS), pages 29114–29125, 2023. 7

    2023

  8. [8]

    High- quality image restoration following human instructions

    Marcos V Conde, Gregor Geigle, and Radu Timofte. High- quality image restoration following human instructions. In Proc. Eur. Conf. Comput. Vis. (ECCV), pages 1–21, 2024. 6, 12

    2024

  9. [9]

    Image restoration via frequency selection.IEEE Trans

    Yuning Cui, Wenqi Ren, Xiaochun Cao, and Alois Knoll. Image restoration via frequency selection.IEEE Trans. Pat- tern Anal. Mach. Intell., 46(2):1093–1108, 2023. 1, 7

    2023

  10. [10]

    Hy- brid frequency modulation network for image restoration

    Yuning Cui, Mingyu Liu, Wenqi Ren, and Alois Knoll. Hy- brid frequency modulation network for image restoration. In Proc. 33rd Int. Joint Conf. Artif. Intell, pages 722–730, 2024. 2

    2024

  11. [11]

    Khan, Alois Knoll, Mubarak Shah, and Fahad Shahbaz Khan

    Yuning Cui, Syed Waqas Zamir, Salman H. Khan, Alois Knoll, Mubarak Shah, and Fahad Shahbaz Khan. Adair: Adaptive all-in-one image restoration via frequency mining and modulation. InInt. Conf. Learn. Represent. (ICLR),

  12. [12]

    FD-GAN: generative adversarial networks with fusion-discriminator for single image dehazing

    Yu Dong, Yihao Liu, He Zhang, Shifeng Chen, and Yu Qiao. FD-GAN: generative adversarial networks with fusion-discriminator for single image dehazing. InProc. AAAI Conf. Artif. Intell. (AAAI), pages 10729–10736, 2020. 1, 7

    2020

  13. [13]

    Kodak Lossless True Color Image Suite,

    Rich Franzen. Kodak Lossless True Color Image Suite,

  14. [14]

    Dy- namic scene deblurring with parameter selective sharing and nested skip connections

    Hongyun Gao, Xin Tao, Xiaoyong Shen, and Jiaya Jia. Dy- namic scene deblurring with parameter selective sharing and nested skip connections. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), pages 3848–3856, 2019. 7

    2019

  15. [15]

    Onerestore: A universal restoration frame- work for composite degradation

    Yu Guo, Yuan Gao, Yuxu Lu, Huilin Zhu, Ryan Wen Liu, and Shengfeng He. Onerestore: A universal restoration frame- work for composite degradation. InProc. Eur. Conf. Comput. Vis. (ECCV), pages 255–272. Springer, 2024. 6, 10

    2024

  16. [16]

    Proto-former: Unified facial landmark detection by prototype transformer.arXiv preprint arXiv:2510.15338, 2025

    Shengkai Hu, Haozhe Qi, Jun Wan, Jiaxing Huang, Lefei Zhang, Hang Sun, and Dacheng Tao. Proto-former: Unified facial landmark detection by prototype transformer.arXiv preprint arXiv:2510.15338, 2025. 1

    arXiv 2025

  17. [17]

    Sin- gle image super-resolution from transformed self-exemplars

    Jia-Bin Huang, Abhishek Singh, and Narendra Ahuja. Sin- gle image super-resolution from transformed self-exemplars. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), pages 5197–5206, 2015. 10

    2015

  18. [18]

    Single target tracking in high-resolution satellite videos: a comprehensive review.Geo-Spatial In- formation Science, 28(3):966–995, 2025

    Xin Huang, Ding Wang, Qiqi Zhu, Ying Zheng, and Qingfeng Guan. Single target tracking in high-resolution satellite videos: a comprehensive review.Geo-Spatial In- formation Science, 28(3):966–995, 2025. 1

    2025

  19. [19]

    Cat-air: Content and task-aware all-in-one image restoration.arXiv preprint arXiv:2503.17915, 2025

    Jiachen Jiang, Tianyu Ding, Ke Zhang, Jinxin Zhou, Tianyi Chen, Ilya Zharkov, Zhihui Zhu, and Luming Liang. Cat-air: Content and task-aware all-in-one image restoration.arXiv preprint arXiv:2503.17915, 2025. 2

    Pith/arXiv arXiv 2025

  20. [20]

    Deblurgan: Blind motion deblurring using conditional adversarial networks

    Orest Kupyn, V olodymyr Budzan, Mykola Mykhailych, Dmytro Mishkin, and Ji ˇr´ı Matas. Deblurgan: Blind motion deblurring using conditional adversarial networks. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), pages 8183–8192, 2018. 7

    2018

  21. [21]

    Benchmarking single- image dehazing and beyond.IEEE Trans

    Boyi Li, Wenqi Ren, Dengpan Fu, Dacheng Tao, Dan Feng, Wenjun Zeng, and Zhangyang Wang. Benchmarking single- image dehazing and beyond.IEEE Trans. Image Process., 28(1):492–505, 2018. 6, 10

    2018

  22. [22]

    All-in-one image restoration for unknown corruption

    Boyun Li, Xiao Liu, Peng Hu, Zhongqin Wu, Jiancheng Lv, and Xi Peng. All-in-one image restoration for unknown corruption. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), pages 17452–17462, 2022. 2, 6, 7, 10, 11

    2022

  23. [23]

    Hybrid agents for image restoration.arXiv preprint arXiv:2503.10120, 2025

    Bingchen Li, Xin Li, Yiting Lu, and Zhibo Chen. Hybrid agents for image restoration.arXiv preprint arXiv:2503.10120, 2025. 2

    Pith/arXiv arXiv 2025

  24. [24]

    Ada4dir: An adaptive model-driven all-in-one image restora- tion network for remote sensing images.Information Fusion, 118:102930, 2025

    Ziyang Lihe, Qiangqiang Yuan, Jiang He, Xianyu Jin, Yi Xiao, Yuzeng Chen, Huanfeng Shen, and Liangpei Zhang. Ada4dir: An adaptive model-driven all-in-one image restora- tion network for remote sensing images.Information Fusion, 118:102930, 2025. 2

    2025

  25. [25]

    Unirestorer: Universal image restoration via adaptively estimating image degrada- tion at proper granularity.arXiv preprint arXiv:2412.20157,

    Jingbo Lin, Zhilu Zhang, Wenbo Li, Renjing Pei, Hang Xu, Hongzhi Zhang, and Wangmeng Zuo. Unirestorer: Universal image restoration via adaptively estimating image degrada- tion at proper granularity.arXiv preprint arXiv:2412.20157,

    Pith/arXiv arXiv

  26. [26]

    Multi-task image restoration guided by robust dino features.arXiv preprint arXiv:2312.01677, 2023

    Xin Lin, Jingtong Yue, Kelvin CK Chan, Lu Qi, Chao Ren, Jinshan Pan, and Ming-Hsuan Yang. Multi-task image restoration guided by robust dino features.arXiv preprint arXiv:2312.01677, 2023. 2

    Pith/arXiv arXiv 2023

  27. [27]

    Tape: Task-agnostic prior embedding for image restoration

    Lin Liu, Lingxi Xie, Xiaopeng Zhang, Shanxin Yuan, Xi- angyu Chen, Wengang Zhou, Houqiang Li, and Qi Tian. Tape: Task-agnostic prior embedding for image restoration. InProc. Eur. Conf. Comput. Vis. (ECCV), pages 447–464,

  28. [28]

    Up- restorer: When unrolling meets prompts for unified image restoration

    Minghao Liu, Wenhan Yang, Jinyi Luo, and Jiaying Liu. Up- restorer: When unrolling meets prompts for unified image restoration. InProc. AAAI Conf. Artif. Intell. (AAAI), pages 5513–5522, 2025. 6

    2025

  29. [29]

    Facial expression recognition with heatmap neighbor contrastive learning.IEEE Trans

    Tong Liu, Jing Li, Jia Wu, Bo Du, Yibing Zhan, Dapeng Tao, and Jun Wan. Facial expression recognition with heatmap neighbor contrastive learning.IEEE Trans. Multimedia,

  30. [30]

    Wm-moe: Weather-aware multi-scale mixture-of-experts for blind adverse weather removal.arXiv preprint arXiv:2303.13739, 2023

    Yulin Luo, Rui Zhao, Xiaobao Wei, Jinwei Chen, Yijie Lu, Shenghao Xie, Tianyu Wang, Ruiqin Xiong, Ming Lu, and Shanghang Zhang. Wm-moe: Weather-aware multi-scale mixture-of-experts for blind adverse weather removal.arXiv preprint arXiv:2303.13739, 2023. 2

    Pith/arXiv arXiv 2023

  31. [31]

    Prores: Exploring degradation-aware visual prompt for universal image restora- tion.arXiv preprint arXiv:2306.13653, 2023

    Jiaqi Ma, Tianheng Cheng, Guoli Wang, Qian Zhang, Xinggang Wang, and Lefei Zhang. Prores: Exploring degradation-aware visual prompt for universal image restora- tion.arXiv preprint arXiv:2306.13653, 2023. 2, 6

    Pith/arXiv arXiv 2023

  32. [32]

    Waterloo exploration database: New challenges for image quality as- sessment models.IEEE Trans

    Kede Ma, Zhengfang Duanmu, Qingbo Wu, Zhou Wang, Hongwei Yong, Hongliang Li, and Lei Zhang. Waterloo exploration database: New challenges for image quality as- sessment models.IEEE Trans. Image Process., 26(2):1004– 1016, 2016. 6, 10

    2016

  33. [33]

    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. InProc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), pages 416–423, 2001. 6, 10

    2001

  34. [34]

    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. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), pages 3883–3891, 2017. 6, 10

    2017

  35. [35]

    Understanding ssim.arXiv preprint arXiv:2006.13846, 2020

    Jim Nilsson and Tomas Akenine-M ¨oller. Understanding ssim.arXiv preprint arXiv:2006.13846, 2020. 6

    Pith/arXiv arXiv 2006

  36. [36]

    Dinov2: Learning robust visual features without supervision

    Maxime Oquab, Timoth ´ee Darcet, Th ´eo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al. Dinov2: Learning robust visual features without supervision. arXiv preprint arXiv:2304.07193, 2023. 2

    Pith/arXiv arXiv 2023

  37. [37]

    Restoring vision in adverse weather conditions with patch-based denoising dif- fusion models.IEEE Trans

    Ozan ¨Ozdenizci and Robert Legenstein. Restoring vision in adverse weather conditions with patch-based denoising dif- fusion models.IEEE Trans. Pattern Anal. Mach. Intell., 45 (8):10346–10357, 2023. 10, 11

    2023

  38. [38]

    Promptir: Prompting for all-in-one blind image restoration

    Vaishnav Potlapalli, Syed Waqas Zamir, Salman Khan, and Fahad Shahbaz Khan. Promptir: Prompting for all-in-one blind image restoration. InAdv. Neural Inf. Process. Syst. (NeurIPS), pages 71275–71293, 2023. 2, 3, 6, 7, 10, 11

    2023

  39. [39]

    Learning transferable visual models from natural language supervi- sion

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervi- sion. InProc. Int. Conf. Mach. Learn. (ICML), pages 8748–

  40. [40]

    Adaptive dy- namic filtering network for image denoising

    Hao Shen, Zhong-Qiu Zhao, and Wandi Zhang. Adaptive dy- namic filtering network for image denoising. InProc. AAAI Conf. Artif. Intell. (AAAI), pages 2227–2235, 2023. 1, 7

    2023

  41. [41]

    Vision transformers for single image dehazing.IEEE Trans

    Yuda Song, Zhuqing He, Hui Qian, and Xin Du. Vision transformers for single image dehazing.IEEE Trans. Image Process., 32:1927–1941, 2023. 1, 7

    1927

  42. [42]

    Baryir: Learning multi-source unified representation in continuous barycenter space for generalizable all-in-one image restora- tion.arXiv preprint arXiv:2505.21637, 2025

    Xiaole Tang, Xiaoyi He, Xiang Gu, and Jian Sun. Baryir: Learning multi-source unified representation in continuous barycenter space for generalizable all-in-one image restora- tion.arXiv preprint arXiv:2505.21637, 2025. 2

    Pith/arXiv arXiv 2025

  43. [43]

    Degradation-aware feature perturbation for all- in-one image restoration

    Xiangpeng Tian, Xiangyu Liao, Xiao Liu, Meng Li, and Chao Ren. Degradation-aware feature perturbation for all- in-one image restoration. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), pages 28165–28175, 2025. 6, 11

    2025

  44. [44]

    Stripformer: Strip transformer for fast image deblurring

    Fu-Jen Tsai, Yan-Tsung Peng, Yen-Yu Lin, Chung-Chi Tsai, and Chia-Wen Lin. Stripformer: Strip transformer for fast image deblurring. InProc. Eur. Conf. Comput. Vis. (ECCV), pages 146–162, 2022. 7

    2022

  45. [45]

    Transweather: Transformer-based restoration of im- ages degraded by adverse weather conditions

    Jeya Maria Jose Valanarasu, Rajeev Yasarla, and Vishal M Patel. Transweather: Transformer-based restoration of im- ages degraded by adverse weather conditions. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), pages 2353–2363, 2022. 6

    2022

  46. [46]

    Precise facial landmark detection by reference heatmap transformer.IEEE Trans

    Jun Wan, Jun Liu, Jie Zhou, Zhihui Lai, Linlin Shen, Hang Sun, Ping Xiong, and Wenwen Min. Precise facial landmark detection by reference heatmap transformer.IEEE Trans. Image Process., 32:1966–1977, 2023. 1

    1966

  47. [47]

    Gewdiff: Geometric enhanced wavelet-based diffu- sion model for hyperspectral image super-resolution.arXiv preprint arXiv:2511.07103, 2025

    Sirui Wang, Jiang He, Natalia Blasco Andreo, and Xiao Xi- ang Zhu. Gewdiff: Geometric enhanced wavelet-based diffu- sion model for hyperspectral image super-resolution.arXiv preprint arXiv:2511.07103, 2025. 2

    arXiv 2025

  48. [48]

    Gridformer: Residual dense transformer with grid struc- ture for image restoration in adverse weather conditions.Int

    Tao Wang, Kaihao Zhang, Ziqian Shao, Wenhan Luo, Bjorn Stenger, Tong Lu, Tae-Kyun Kim, Wei Liu, and Hongdong Li. Gridformer: Residual dense transformer with grid struc- ture for image restoration in adverse weather conditions.Int. J. Comput. Vis., 132(10):4541–4563, 2024. 6

    2024

  49. [49]

    M2restore: Mixture-of-experts- based mamba-cnn fusion framework for all-in-one image restoration.arXiv preprint arXiv:2506.07814, 2025

    Yongzhen Wang, Yongjun Li, Zhuoran Zheng, Xiao-Ping Zhang, and Mingqiang Wei. M2restore: Mixture-of-experts- based mamba-cnn fusion framework for all-in-one image restoration.arXiv preprint arXiv:2506.07814, 2025. 2, 3

    Pith/arXiv arXiv 2025

  50. [50]

    Deep retinex decomposition for low-light enhancement

    Chen Wei, Wenjing Wang, Wenhan Yang, and Jiaying Liu. Deep retinex decomposition for low-light enhancement. In Proc. Brit. Mach. Vis. Conf. (BMVC), 2018. 6, 10

    2018

  51. [51]

    Dswinir: Rethinking window-based attention for image restoration, 2025

    Gang Wu, Junjun Jiang, Kui Jiang, Xianming Liu, and Liqiang Nie. Dswinir: Rethinking window-based attention for image restoration, 2025. 2

    2025

  52. [52]

    Uretinex-net: Retinex-based deep unfolding network for low-light image enhancement

    Wenhui Wu, Jian Weng, Pingping Zhang, Xu Wang, Wen- han Yang, and Jianmin Jiang. Uretinex-net: Retinex-based deep unfolding network for low-light image enhancement. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), pages 5901–5910, 2022. 1, 2, 7

    2022

  53. [53]

    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. In Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), pages 13095–13105, 2023. 7

    2023

  54. [54]

    Occlusion-embedded hybrid transformer for light field super-resolution

    Zeyu Xiao, Zhuoyuan Li, and Wei Jia. Occlusion-embedded hybrid transformer for light field super-resolution. InAAAI Conf. Artif. Intell. (AAAI), pages 8700–8708, 2025. 1

    2025

  55. [55]

    Low-light image enhancement via structure modeling and guidance

    Xiaogang Xu, Ruixing Wang, and Jiangbo Lu. Low-light image enhancement via structure modeling and guidance. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), pages 9893–9903, 2023. 7 19

    2023

  56. [56]

    Language- driven all-in-one adverse weather removal

    Hao Yang, Liyuan Pan, Yan Yang, and Wei Liang. Language- driven all-in-one adverse weather removal. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), pages 24902–24912, 2024. 2

    2024

  57. [57]

    Deep joint rain detection and removal from a single image

    Wenhan Yang, Robby T Tan, Jiashi Feng, Jiaying Liu, Zong- ming Guo, and Shuicheng Yan. Deep joint rain detection and removal from a single image. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), pages 1357–1366,

  58. [58]

    Neural degradation representation learning for all-in-one image restoration.IEEE Trans

    Mingde Yao, Ruikang Xu, Yuanshen Guan, Jie Huang, and Zhiwei Xiong. Neural degradation representation learning for all-in-one image restoration.IEEE Trans. Image Pro- cess., 33:5408–5423, 2024. 6

    2024

  59. [59]

    Wave-vit: Unifying wavelet and transformers for vi- sual representation learning

    Ting Yao, Yingwei Pan, Yehao Li, Chong-Wah Ngo, and Tao Mei. Wave-vit: Unifying wavelet and transformers for vi- sual representation learning. InProc. Eur. Conf. Comput. Vis. (ECCV), pages 328–345. Springer, 2022. 3

    2022

  60. [60]

    Uncertainty guided multi-scale residual learning-using a cycle spinning cnn for single image de-raining

    Rajeev Yasarla and Vishal M Patel. Uncertainty guided multi-scale residual learning-using a cycle spinning cnn for single image de-raining. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), pages 8405–8414, 2019. 7

    2019

  61. [61]

    Multi-expert adaptive selection: Task-balancing for all-in- one image restoration.IEEE Trans

    Xiaoyan Yu, Shen Zhou, Huafeng Li, and Liehuang Zhu. Multi-expert adaptive selection: Task-balancing for all-in- one image restoration.IEEE Trans. Circuits Syst. Video Technol., 2024. 2, 3

    2024

  62. [62]

    Com- plexity experts are task-discriminative learners for any image restoration

    Eduard Zamfir, Zongwei Wu, Nancy Mehta, Yuedong Tan, Danda Pani Paudel, Yulun Zhang, and Radu Timofte. Com- plexity experts are task-discriminative learners for any image restoration. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), pages 12753–12763, 2025. 2, 6, 11, 12

    2025

  63. [63]

    Learning enriched features for real image restoration and enhancement

    Syed Waqas Zamir, Aditya Arora, Salman Khan, Munawar Hayat, Fahad Shahbaz Khan, Ming-Hsuan Yang, and Ling Shao. Learning enriched features for real image restoration and enhancement. InProc. Eur. Conf. Comput. Vis. (ECCV), pages 492–511, 2020. 1, 7

    2020

  64. [64]

    Khan, Mu- nawar Hayat, Fahad Shahbaz Khan, Ming-Hsuan Yang, and Ling Shao

    Syed Waqas Zamir, Aditya Arora, Salman H. Khan, Mu- nawar Hayat, Fahad Shahbaz Khan, Ming-Hsuan Yang, and Ling Shao. Multi-stage progressive image restoration. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), pages 14816–14826, 2021. 2, 7

    2021

  65. [65]

    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. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), pages 5728–5739, 2022. 1, 7

    2022

  66. [66]

    Learning enriched features for fast image restoration and enhancement.IEEE Trans

    Syed Waqas Zamir, Aditya Arora, Salman Khan, Munawar Hayat, Fahad Shahbaz Khan, Ming-Hsuan Yang, and Ling Shao. Learning enriched features for fast image restoration and enhancement.IEEE Trans. Pattern Anal. Mach. Intell., 45(2):1934–1948, 2022. 1, 2, 7

    1934

  67. [67]

    Ingredient-oriented multi-degradation learning for image restoration

    Jinghao Zhang, Jie Huang, Mingde Yao, Zizheng Yang, Hu Yu, Man Zhou, and Feng Zhao. Ingredient-oriented multi-degradation learning for image restoration. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), pages 5825–5835, 2023. 6, 7

    2023

  68. [68]

    A tale of two features: Stable diffusion complements dino for zero-shot semantic correspondence

    Junyi Zhang, Charles Herrmann, Junhwa Hur, Luisa Pola- nia Cabrera, Varun Jampani, Deqing Sun, and Ming-Hsuan Yang. A tale of two features: Stable diffusion complements dino for zero-shot semantic correspondence. InAdv. Neural Inf. Process. Syst. (NeurIPS), 2024. 2

    2024

  69. [69]

    Beyond a gaussian denoiser: Residual learning of deep cnn for image denoising.IEEE Trans

    Kai Zhang, Wangmeng Zuo, Yunjin Chen, Deyu Meng, and Lei Zhang. Beyond a gaussian denoiser: Residual learning of deep cnn for image denoising.IEEE Trans. Image Process., 26(7):3142–3155, 2017. 1, 7

    2017

  70. [70]

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

    Kai Zhang, Wangmeng Zuo, and Lei Zhang. Ffdnet: Toward a fast and flexible solution for cnn-based image denoising. IEEE Trans. Image Process., 27(9):4608–4622, 2018. 1, 7

    2018

  71. [71]

    Efficient deweahter mixture- of-experts with uncertainty-aware feature-wise linear mod- ulation

    Rongyu Zhang, Yulin Luo, Jiaming Liu, Huanrui Yang, Zhen Dong, Denis Gudovskiy, Tomoyuki Okuno, Yohei Nakata, Kurt Keutzer, Yuan Du, et al. Efficient deweahter mixture- of-experts with uncertainty-aware feature-wise linear mod- ulation. InAAAI Conf. Artif. Intell. (AAAI), pages 16812– 16820, 2024. 2

    2024

  72. [72]

    Perceive-ir: Learning to perceive degra- dation better for all-in-one image restoration.IEEE Trans

    Xu Zhang, Jiaqi Ma, Guoli Wang, Qian Zhang, Huan Zhang, and Lefei Zhang. Perceive-ir: Learning to perceive degra- dation better for all-in-one image restoration.IEEE Trans. Image Process., pages 1–1, 2025. 2, 6, 7

    2025

  73. [73]

    Seeing the unseen: A fre- quency prompt guided transformer for image restoration

    Shihao Zhou, Jinshan Pan, Jinglei Shi, Duosheng Chen, Lishen Qu, and Jufeng Yang. Seeing the unseen: A fre- quency prompt guided transformer for image restoration. InProc. Eur. Conf. Comput. Vis. (ECCV), pages 246–264. Springer, 2024. 2

    2024

  74. [74]

    Learn- ing weather-general and weather-specific features for im- age restoration under multiple adverse weather conditions

    Yurui Zhu, Tianyu Wang, Xueyang Fu, Xuanyu Yang, Xin Guo, Jifeng Dai, Yu Qiao, and Xiaowei Hu. Learn- ing weather-general and weather-specific features for im- age restoration under multiple adverse weather conditions. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), pages 21747–21758, 2023. 10, 11 20

    2023

  75. [1999]

    Available:http : / / r0k

    [Online]. Available:http : / / r0k . us / graphics/kodak/[Accessed: Oct. 24, 2021]. 6

    2021