Pith. sign in

REVIEW 3 major objections 8 minor 46 references

ReDiffuse: Rotation Equivariant Diffusion Model for Multi-focus Image Fusion

T0 review · 3 major / 8 minor · reviewed 2026-07-13 · grok-4.5

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

Pith's one-line read Embedding rotation equivariance into a diffusion U-Net stops defocus from warping fused multi-focus images.

desk verdict Solid subfield methods paper: end-to-end rotation-equivariant diffusion U-Net for MFIF with real equivariance-error analysis and consistent gains; theory is weakest exactly on hard defocus, but the empirical package still deserves referees. read the letter →

arxiv 2603.21129 v1 pith:JPL4EER5 submitted 2026-03-22 cs.CV

Bo Li , Tingting Bao , Lingling Zhang , Weiping Fu , Yaxian Wang , Jun Liu This is my paper · ORCID
classification cs.CV
keywords RotationEquivariantDiffusionModelMulti-focusImageFusionEquivarianceErrorAnalysisU-NetB-ConvDefocusBlur
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

Multi-focus image fusion must combine several differently focused shots into one sharp all-in-focus picture. Diffusion models can generate high-fidelity results, yet defocus blur warps edges and textures, so the network often invents artifacts while denoising. The paper argues that the missing inductive bias is rotation equivariance: if a pattern is rotated, the network’s response must rotate with it, preserving orientation and local isotropic symmetry. ReDiffuse rebuilds every block of the diffusion U-Net (convolutions, down-sampling, up-sampling, normalization) so that the whole pipeline is approximately rotation-equivariant end-to-end. A sequence of theorems bounds the residual equivariance error by the mesh size and shows that the error is exactly zero for the common 90-degree rotation group. On four standard benchmarks the method improves six quality metrics by 0.28–6.64 percent while cutting parameters from 26.91 M to 7.55 M. A sympathetic reader cares because the same geometric prior can be dropped into other diffusion fusion models and immediately raises their scores, and because the theory supplies a concrete reason why the prior works.

What carries the argument

Rotation-equivariant (Rot-E) U-Net built from B-Conv filters, equivariant max-pooling / bilinear up-sampling, and group normalization; Theorems 3.1–3.4 and Corollaries 3.5–3.6 bound its intrinsic equivariance error by mesh size δ and prove exact equivariance when m = 4.

What would settle it

Replace every Rot-E block with ordinary convolutions while keeping identical capacity and training; if the six fusion metrics on Lytro, MFFW, MFI-WHU and Road-MF no longer improve and the measured equivariance-error maps no longer collapse to near zero under 90/180/270 rotations, the central claim fails.

Watch

Extended reading notes

Core claim

Defocus blur destroys reliable orientation cues inside a diffusion denoiser; forcing the entire U-Net to be rotation-equivariant restores those cues, so fused images keep the original directional consistency of textures and edges. For the discrete 90-degree group the equivariance error is theoretically zero, and experiments confirm both smaller residual maps and higher fusion metrics.

Load-bearing premise

The error bounds assume that latent continuous image and filter maps have uniformly bounded gradients and Hessians; under strong defocus those local derivatives become unreliable, which is exactly where the method can still fail.

Editorial extensions

If this is right

  • Existing diffusion MFIF backbones can be upgraded by swapping in Rot-E modules and should inherit both better metrics and a roughly 3 imes parameter reduction.
  • For the standard four-fold rotation group the theoretical equivariance error vanishes, so residual orientation artifacts can no longer be blamed on the architecture.
  • The same construction supplies a practical recipe for any diffusion image-restoration task whose ground-truth structures are approximately rotation-stable.
  • Failure under severe defocus spread points to non-local correlation modules as the next architectural addition needed to keep the equivariance guarantee useful.

Reading between the lines

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

  • Because the equivariance guarantee is discrete and mesh-size dependent, continuous or arbitrary-angle rotations still leave a residual error that may matter for camera-shake or free-form viewpoint changes.
  • The same Rot-E blocks could be inserted into multi-modal fusion pipelines (infrared-visible, medical) where orientation consistency is equally fragile under blur or sensor misalignment.
  • If non-local attention is later combined with the Rot-E U-Net, the failure case on strong defocus spread may disappear without sacrificing the zero-error corollary for m = 4.
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 / 8 minor

Summary. The paper proposes ReDiffuse, a diffusion model for multi-focus image fusion (MFIF) whose U-Net backbone is redesigned for approximate end-to-end rotation equivariance via B-Conv (Rot-E) layers, equivariant residual blocks, and redesigned down/upsampling. The motivation is that defocus blur warps edges and textures, so an equivariant inductive bias should better preserve orientation consistency and local isotropic symmetry during conditional denoising. The authors give equivariance-error bounds for max-pooling, bilinear upsampling, and group normalization (Theorems 3.1–3.3), an O(δ) bound for the full network (Theorem 3.4), exact equivariance for the discrete m=4 group (Corollary 3.5), and a continuous-angle bound (Corollary 3.6). Empirically, ReDiffuse is trained on Real-MFF and compared on Lytro, MFFW, MFI-WHU, and Road-MF against 11 baselines across six no-reference metrics, with ablations, plug-in tests on three other diffusion fusion models, equivariance-error maps, and a failure-case discussion. Parameter count is reduced from 26.91M to 7.55M via group parameter sharing.

Significance. If the empirical gains hold under fair comparison, the work is a solid, practically useful contribution to MFIF: it couples a modern diffusion fusion pipeline with a carefully constructed rotation-equivariant U-Net, reports consistent improvements on four public benchmarks, and ships code. The lightweight parameter sharing is a concrete engineering benefit. The theoretical analysis (module-wise bounds plus full-network corollaries) is a genuine strength relative to most fusion papers, even though the regularity assumptions limit how far the guarantees extend under severe defocus. The plug-in generalization experiment (Table 2) and explicit failure-case analysis (§4.6) increase credibility. Overall significance is incremental-to-moderate for the MFIF/diffusion community rather than a foundational advance in equivariant deep learning, but it is a well-scoped and useful application paper.

major comments (3)
  1. §3.3 (Eq. 20), Theorems 3.1–3.4, and Corollaries 3.5–3.6 vs. the central claim in the abstract/§1: the equivariance-error theory controls discrete error by mesh size δ only under uniform bounds on values, gradients, and Hessians of latent continuous image/filter maps. Defocus is precisely a regime that can make those local derivatives unreliable. The paper’s own §4.6 failure case on MFFW attributes incorrect structural selection to strong defocus spread and explicitly links it to larger G0/H0 in Corollary 3.6. The theory therefore provides the weakest formal support exactly where the introduction claims the largest benefit. This does not invalidate the empirical results, but the abstract and §1 currently overstate what the theory guarantees under defocus. Please restate the theoretical claims as: (i) exact equivariance for m=4 under the stated regularity assumptions / discrete grid align
  2. §3.2–3.3 and Figure 4: the analysis establishes (approximate) equivariance of the noise-predictor U-Net modules, but the MFIF pipeline is an iterative reverse process (Eqs. 4–6) conditioned on two source images IA, IB concatenated with Ft. For the end-to-end claim to be load-bearing, please state explicitly that (a) rotating (IA, IB, Ft) jointly induces the corresponding rotation of εθ, and (b) the deterministic reverse step (Eq. 6) preserves equivariance because it is an equivariant function of Ft and εθ. A short proposition or remark would close the gap between “Rot-E U-Net is equivariant” and “the full ReDiffuse fusion map is equivariant.” Without this, “end-to-end rotation equivariance” is slightly stronger than what is proved in the main text.
  3. §4.1 Implementation / Table 1: fairness of the multi-method comparison needs a clearer protocol statement. The paper says competing methods use “official provided code and weights,” while ReDiffuse is trained 10,000 epochs on Real-MFF. If several baselines were trained on different data, different pair counts, or released only for other fusion settings, the 0.28–6.64% average margins over the second-best method are hard to interpret as pure architectural gains. Please add a short table or paragraph listing, for each baseline, training data, whether weights were used as released or retrained, and input resolution. If retraining all diffusion baselines on Real-MFF under the same schedule is infeasible, at least retrain the closest non-equivariant backbone (the “w/o rotation equivariance” model already in Table 3) and report that as the primary controlled comparison, with official-weight ba
minor comments (8)
  1. Figure 1 caption and §1: “non-local directional similarity” and “local isotropic symmetry” are illustrated but not operationalized. A one-sentence definition or pointer to how these are measured (even qualitatively) would help readers who are not already convinced by the red/blue boxes.
  2. Table 1, Lytro: TC-MoA exceeds ReDiffuse on QMI (0.993 vs 0.982). The text says ReDiffuse is best “across nearly all” metrics; please acknowledge the few second-place entries explicitly when claiming average improvements.
  3. §3.2: T=2000 is stated without justification relative to FusionDiff or other diffusion MFIF work. A brief note on sampling steps at inference (and any DDIM-style acceleration) belongs in §4.1 or Appendix B.
  4. Related Work §2.2: the claim that “theoretical guarantees for enforcing rotation equivariance within diffusion models remain largely unexplored” should cite any concurrent equivariant-diffusion work outside MFIF (even if not fusion-specific) to avoid overclaiming novelty of the theory setting.
  5. Notation: π^I_R̃ vs π^F_R̃ and the composition notation [·] in Eq. (9) are dense; a short example for m=4 on a single feature map in Appendix A.1 would improve accessibility.
  6. Figure 8: “after 1,000 training epochs” is a mid-training snapshot. Please also report final-epoch equivariance error (mean/max over the test set) so the maps can be tied to the deployed model.
  7. Typos / polish: “ReDiffusetheoretically” (Fig. 4 caption); “o90” labels in Fig. 1; inconsistent spacing around citations (e.g., “methods [2, 26, 38?]”); “MS-SSIM:0.970” style in Fig. 2 could use a space after the colon.
  8. §4.2: runtime is deferred to Appendix C.1; a one-line statement in the main text (e.g., relative FPS vs FusionDiff) would help readers assess the cost of equivariance.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: equivariance is an architectural inductive bias with independent error bounds under stated regularity assumptions, and fusion metrics are external benchmarks.

full rationale

The paper's derivation chain does not reduce any claimed prediction or first-principles result to its own inputs by construction. The forward/reverse diffusion processes and noise-prediction loss (Eqs. 1–7) are the standard DDPM formulation conditioned on source images; nothing is fitted to the six fusion metrics later reported. Rotation equivariance is imposed by redesigning U-Net modules (B-Conv, Rot-E down/upsampling, GN) and is analyzed via Theorems 3.1–3.4 and Corollaries 3.5–3.6, which bound discrete equivariance error by mesh size δ under explicit uniform bounds on latent continuous maps and filters (Eq. 20). Those bounds are mathematical consequences of the stated regularity assumptions, not tautological restatements of the empirical scores. B-Conv/F-Conv/G-Conv are cited as prior filter-parameterization tools (external authors) used as components; no uniqueness theorem or load-bearing premise is imported solely from overlapping self-citations. Empirical gains are measured against independent public MFIF datasets and standard metrics (Qabf, QMI, etc.) with official competitor code. The failure-case discussion in §4.6 correctly notes that strong defocus can violate the regularity assumptions, which is a correctness/scope limitation, not circularity. The architecture, theory, and evaluation are therefore self-contained against external benchmarks.

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

The load-bearing content is an architectural inductive bias plus regularity-based error bounds and empirical fusion scores. Free parameters are standard training/architecture choices. Axioms mix standard diffusion math, prior equivariant-convolution results, and paper-specific continuous-map regularity used to bound discrete equivariance error. The main invented entity is the ReDiffuse Rot-E U-Net composition itself, not a new physical object.

free parameters (4)
  • rotation group order m
    Set to 4 for zero discrete equivariance error (Corollary 3.5) and 1/m parameter sharing; choice is design, not derived from data uniqueness.
  • diffusion steps T and noise schedule β_t
    T=2000 and the standard DDPM-style schedule are chosen hyperparameters of the generative process.
  • U-Net width / GroupNorm groups / training schedule
    Init channels 64, GN groups 16, 10k epochs, batch 32, Adam lr 0.0002 with decay—hand-set training configuration that affects reported metrics.
  • mesh size δ and filter support p in equivariance bounds
    Appear in Theorems 3.1–3.4 as discretization parameters controlling C1δ error; not measured from images but assumed small enough for the bound narrative.
assumptions (5)
  • domain assumption Standard conditional DDPM forward/reverse process and L2 noise-prediction training (Eqs. 1–7), with deterministic mean-only sampling for MFIF.
    Taken from diffusion literature and prior MFIF diffusion work (e.g., FusionDiff); not re-proved.
  • domain assumption Group convolution / B-Conv and pointwise SiLU preserve rotation equivariance; analysis can focus on pool, bilinear upsample, and GN.
    Cited from prior equivariant CNN results ([6], [30], [34]); paper builds on them rather than re-deriving filter equivariance.
  • ad hoc to paper Latent continuous image and filter maps have bounded values, gradients, and Hessians (Eq. 20), enabling O(δ) equivariance-error bounds for the full network.
    Essential for Theorems 3.1–3.4 and Corollaries 3.5–3.6; not independently verified on defocused MFIF imagery and is strained in the failure case.
  • domain assumption Rotation equivariance is an appropriate geometric prior for MFIF because non-local self-similarity and low-rank structure remain stable under rotation (Fig. 3 narrative).
    Motivates the architecture; plausible but not uniquely forced—other symmetries or non-local modules could be alternatives.
  • domain assumption No-reference fusion metrics (Qabf, QMI, QG, QP, QE) and source-averaged MS-SSIM are adequate proxies for fusion quality without ground-truth all-in-focus images.
    Standard in MFIF evaluation ([19], [44], [45]); still a measurement axiom for the empirical claim.
invented entities (1)
  • ReDiffuse Rot-E U-Net (end-to-end rotation-equivariant diffusion backbone for MFIF)
    purpose: Implement approximate end-to-end rotation equivariance during conditional denoising to reduce orientation-warping fusion artifacts.
    Composition of B-Conv, Rot-E down/upsampling, ResBlocks with time embedding, and GN+SiLU stages is the paper’s proposed system; independent evidence is empirical tables and equivariance-error maps, not an external physical prediction.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ReDiffuse: Rotation Equivariant Diffusion Model for Multi-focus Image Fusion." pith.science (2026). https://pith.science/paper/JPL4EER5

@misc{pith2026260321129,
  author       = {Pith},
  title        = {Pith review of: ReDiffuse: Rotation Equivariant Diffusion Model for Multi-focus Image Fusion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JPL4EER5}},
  note         = {Machine review of arXiv:2603.21129}
}
read the original abstract

Diffusion models have achieved impressive performance on multi-focus image fusion (MFIF). However, a key challenge in applying diffusion models to the ill-posed MFIF problem is that defocus blur can make common symmetric geometric structures (e.g., textures and edges) appear warped and deformed, often leading to unexpected artifacts in the fused images. Therefore, embedding rotation equivariance into diffusion networks is essential, as it enables the fusion results to faithfully preserve the original orientation and structural consistency of geometric patterns underlying the input images. Motivated by this, we propose ReDiffuse, a rotation-equivariant diffusion model for MFIF. Specifically, we carefully construct the basic diffusion architectures to achieve end-to-end rotation equivariance. We also provide a rigorous theoretical analysis to evaluate its intrinsic equivariance error, demonstrating the validity of embedding equivariance structures. ReDiffuse is comprehensively evaluated against various MFIF methods across four datasets (Lytro, MFFW, MFI-WHU, and Road-MF). Results demonstrate that ReDiffuse achieves competitive performance, with improvements of 0.28-6.64\% across six evaluation metrics. The code is available at https://github.com/MorvanLi/ReDiffuse.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references

  1. [1]

    Haowen Bai, Zixiang Zhao, Jiangshe Zhang, Yichen Wu, Lilun Deng, Yukun Cui, Baisong Jiang, and Shuang Xu. 2025. Refusion: Learning image fusion from reconstruction with learnable loss via meta-learning.International Journal of Computer Vision133, 5 (2025), 2547–2567

    2025

  2. [2]

    Bing Cao, Xingxin Xu, Pengfei Zhu, Qilong Wang, and Qinghua Hu. 2024. Con- ditional controllable image fusion.Advances in Neural Information Processing Systems37 (2024), 120311–120335

    2024

  3. [3]

    Chunyang Cheng, Tianyang Xu, Zhenhua Feng, Xiaojun Wu, Zhangyong Tang, Hui Li, Zeyang Zhang, Sara Atito, Muhammad Awais, and Josef Kittler. 2025. One Model for ALL: Low-Level Task Interaction Is a Key to Task-Agnostic Image Fusion. InProceedings of the Computer Vision and Pattern Recognition Conference. 28102–28112

    2025

  4. [4]

    Taco Cohen and Max Welling. 2016. Group equivariant convolutional networks. InInternational conference on machine learning. PMLR, 2990–2999

    2016

  5. [5]

    Xin Deng, Jingyi Xu, Fangyuan Gao, Xiancheng Sun, and Mai Xu. 2024. DeepM2CDL: Deep Multi-Scale Multi-Modal Convolutional Dictionary Learning Network.IEEE Transactions on Pattern Analysis and Machine Intelligence46, 5 (2024), 2770–2787

    2024

  6. [6]

    Jiahong Fu, Qi Xie, Deyu Meng, and Zongben Xu. 2024. Rotation Equivariant Proximal Operator for Deep Unfolding Methods in Image Restoration.IEEE Transactions on Pattern Analysis and Machine Intelligence46, 10 (2024), 6577– 6593

    2024

  7. [7]

    Xiaopeng Guo, Rencan Nie, Jinde Cao, Dongming Zhou, Liye Mei, and Kangjian He. 2019. FuseGAN: Learning to fuse multi-focus image via conditional gener- ative adversarial network.IEEE Transactions on Multimedia21, 8 (2019), 1982– 1996

    2019

  8. [8]

    Dan He, Weisheng Li, Guofen Wang, Yuping Huang, and Shiqiang Liu. 2025. DM- FNet: Unified Multimodal Medical Image Fusion via Diffusion Process-Trained Encoder-Decoder.IEEE Transactions on Multimedia27 (2025), 9415–9428

    2025

Show all 46 references
  1. [9]

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. 2020. Denoising diffusion probabilistic models.Advances in neural information processing systems33 (2020), 6840–6851

    2020

  2. [10]

    Xingyu Hu, Junjun Jiang, Xianming Liu, and Jiayi Ma. 2023. ZMFF: Zero-shot multi-focus image fusion.Information Fusion92 (2023), 127–138

    2023

  3. [11]

    Jun Huang, Zhuliang Le, Yong Ma, Xiaoguang Mei, and Fan Fan. 2020. A genera- tive adversarial network with adaptive constraints for multi-focus image fusion. Neural Computing and Applications32, 18 (2020), 15119–15129

    2020

  4. [12]

    Xin Jin, Pengcheng Zhu, Dongjian Yu, Michal Wozniak, Qian Jiang, Puming Wang, and Wei Zhou. 2025. Combining depth and frequency features with Mamba for multi-focus image fusion.Information Fusion(2025), 103355

    2025

  5. [13]

    Bo Li, Lingling Zhang, Tingting Bao, Yunkuo Lei, Xiaoqing Zhang, and Jun Liu. 2026. When Multi-Focus Image Fusion Meets Nonlinear Spiking Neural P Systems.IEEE Transactions on Multimedia28 (2026), 545–560

    2026

  6. [14]

    Jinxing Li, Xiaobao Guo, Guangming Lu, Bob Zhang, Yong Xu, Feng Wu, and David Zhang. 2020. DRPL: Deep regression pair learning for multi-focus image fusion.IEEE Transactions on Image Processing29 (2020), 4816–4831

    2020

  7. [15]

    Mining Li, Ronghao Pei, Tianyou Zheng, Yang Zhang, and Weiwei Fu. 2024. FusionDiff: Multi-focus image fusion using denoising diffusion probabilistic models.Expert Systems with Applications238 (2024), 121664

    2024

  8. [16]

    Xilai Li, Xiaosong Li, Haishu Tan, and Jinyang Li. 2024. SAMF: small-area- aware multi-focus image fusion for object detection. InICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 3845–3849

    2024

  9. [17]

    Hanze Liu, Jiahong Fu, Qi Xie, and Deyu Meng. 2025. Rotation-equivariant self-supervised method in image denoising. InProceedings of the Computer Vision and Pattern Recognition Conference. 12720–12730

    2025

  10. [18]

    Risheng Liu, Zhu Liu, Jinyuan Liu, and Xin Fan. 2021. Searching a hierarchi- cally aggregated fusion architecture for fast multi-modality image fusion. In Proceedings of the 29th ACM International Conference on Multimedia. 1600–1608

    2021

  11. [19]

    Yu Liu, Zhengzheng Qi, Juan Cheng, and Xun Chen. 2024. Rethinking the Effectiveness of Objective Evaluation Metrics in Multi-focus Image Fusion: A Statistic-based Approach.IEEE Transactions on Pattern Analysis and Machine Intelligence46, 8 (2024), 5806–5819

    2024

  12. [20]

    Yu Liu, Lei Wang, Juan Cheng, and Xun Chen. 2021. Multiscale feature interactive network for multifocus image fusion.IEEE Transactions on Instrumentation and Measurement70 (2021), 1–16

    2021

  13. [21]

    Yu Liu, Lei Wang, Juan Cheng, Chang Li, and Xun Chen. 2020. Multi-focus image fusion: A survey of the state of the art.Information Fusion64 (2020), 71–91

    2020

  14. [22]

    Jiayi Ma, Linfeng Tang, Fan Fan, Jun Huang, Xiaoguang Mei, and Yong Ma. 2022. SwinFusion: Cross-domain long-range learning for general image fusion via swin transformer.IEEE/CAA Journal of Automatica Sinica9, 7 (2022), 1200–1217

    2022

  15. [23]

    Mansour Nejati, Shadrokh Samavi, and Shahram Shirani. 2015. Multi-focus image fusion using dictionary-based sparse representation.Information Fusion 25 (2015), 72–84

    2015

  16. [24]

    Yuhui Quan, Xi Wan, Zitao Tang, Jinxiu Liang, and Hui Ji. 2025. Multi-Focus Image Fusion via Explicit Defocus Blur Modelling. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 39. 6657–6665

    2025

  17. [25]

    Jane Wang, and Xun Chen

    Yu Shi, Yu Liu, Juan Cheng, Z. Jane Wang, and Xun Chen. 2025. VDMUFusion: A Versatile Diffusion Model-Based Unsupervised Framework for Image Fusion. IEEE Transactions on Image Processing34 (2025), 441–454

    2025

  18. [26]

    Linfeng Tang, Chunyu Li, and Jiayi Ma. 2026. Mask-DiFuser: A Masked Diffusion Model for Unified Unsupervised Image Fusion.IEEE Transactions on Pattern Analysis and Machine Intelligence48, 1 (2026), 591–608

    2026

  19. [27]

    Xiaolong Wang, Ross Girshick, Abhinav Gupta, and Kaiming He. 2018. Non-local neural networks. InProceedings of the IEEE conference on computer vision and pattern recognition. 7794–7803

    2018

  20. [28]

    Yancheng Wang, Yang Xiao, Junyi Lu, Bo Tan, Zhiguo Cao, Zhenjun Zhang, and Joey Tianyi Zhou. 2021. Discriminative multi-view dynamic image fusion for cross-view 3-D action recognition.IEEE Transactions on Neural Networks and Learning Systems33, 10 (2021), 5332–5345

    2021

  21. [29]

    Zeyu Wang, Xiongfei Li, Libo Zhao, Haoran Duan, Shidong Wang, Hao Liu, and Xiaoli Zhang. 2023. When multi-focus image fusion networks meet traditional edge-preservation technology.International Journal of Computer Vision131, 10 (2023), 2529–2552

    2023

  22. [30]

    Maurice Weiler and Gabriele Cesa. 2019. General e (2)-equivariant steerable cnns.Advances in neural information processing systems32 (2019)

    2019

  23. [31]

    Maurice Weiler, Fred A Hamprecht, and Martin Storath. 2018. Learning steerable filters for rotation equivariant cnns. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 849–858

    2018

  24. [32]

    Daniel E Worrall, Stephan J Garbin, Daniyar Turmukhambetov, and Gabriel J Brostow. 2017. Harmonic networks: Deep translation and rotation equivariance. InProceedings of the IEEE conference on computer vision and pattern recognition. 5028–5037

    2017

  25. [33]

    Bin Xiao, Bocheng Xu, Xiuli Bi, and Weisheng Li. 2020. Global-feature encoding U-Net (GEU-Net) for multi-focus image fusion.IEEE Transactions on Image Processing30 (2020), 163–175

    2020

  26. [34]

    Qi Xie, Jiahong Fu, Zongben Xu, and Deyu Meng. 2025. Rotation Equivariant Arbitrary-Scale Image Super-Resolution.IEEE Transactions on Pattern Analysis and Machine Intelligence47, 12 (2025), 11382–11399

    2025

  27. [35]

    Qi Xie, Qian Zhao, Zongben Xu, and Deyu Meng. 2023. Fourier Series Expansion Based Filter Parametrization for Equivariant Convolutions.IEEE Transactions on Pattern Analysis and Machine Intelligence45, 4 (2023), 4537–4551

    2023

  28. [36]

    Han Xu, Jiayi Ma, Junjun Jiang, Xiaojie Guo, and Haibin Ling. 2020. U2Fusion: A unified unsupervised image fusion network.IEEE Transactions on Pattern Analysis and Machine Intelligence44, 1 (2020), 502–518

    2020

  29. [37]

    Shuang Xu, Lizhen Ji, Zhe Wang, Pengfei Li, Kai Sun, Chunxia Zhang, and Jiangshe Zhang. 2020. Towards reducing severe defocus spread effects for multi- focus image fusion via an optimization based strategy.IEEE Transactions on Computational Imaging6 (2020), 1561–1570

    2020

  30. [38]

    Xunpeng Yi, Linfeng Tang, Hao Zhang, Han Xu, and Jiayi Ma. 2024. Diff-IF: Multi-modality image fusion via diffusion model with fusion knowledge prior. Information Fusion110 (2024), 102450

    2024

  31. [39]

    Hao Zhang, Zhuliang Le, Zhenfeng Shao, Han Xu, and Jiayi Ma. 2021. MFF-GAN: An unsupervised generative adversarial network with adaptive and gradient 9 joint constraints for multi-focus image fusion.Information Fusion66 (2021), 40–53

    2021

  32. [40]

    Hao Zhang, Han Xu, Yang Xiao, Xiaojie Guo, and Jiayi Ma. 2020. Rethinking the image fusion: A fast unified image fusion network based on proportional maintenance of gradient and intensity. InProceedings of the AAAI conference on artificial intelligence, Vol. 34. 12797–12804

    2020

  33. [41]

    Juncheng Zhang, Qingmin Liao, Shaojun Liu, Haoyu Ma, Wenming Yang, and Jing-Hao Xue. 2020. Real-MFF: A large realistic multi-focus image dataset with ground truth.Pattern Recognition Letters138 (2020), 370–377

    2020

  34. [42]

    Xingchen Zhang. 2021. Deep learning-based multi-focus image fusion: A survey and a comparative study.IEEE Transactions on Pattern Analysis and Machine Intelligence44, 9 (2021), 4819–4838

    2021

  35. [43]

    Yu Zhang, Yu Liu, Peng Sun, Han Yan, Xiaolin Zhao, and Li Zhang. 2020. IFCNN: A general image fusion framework based on convolutional neural network. Information Fusion54 (2020), 99–118

    2020

  36. [44]

    Zixiang Zhao, Haowen Bai, Jiangshe Zhang, Yulun Zhang, Shuang Xu, Zudi Lin, Radu Timofte, and Luc Van Gool. 2023. Cddfuse: Correlation-driven dual-branch feature decomposition for multi-modality image fusion. InProceedings of the IEEE/CVF conference on computer vision and patt...

    2023

  37. [45]

    Zixiang Zhao, Haowen Bai, Yuanzhi Zhu, Jiangshe Zhang, Shuang Xu, Yulun Zhang, Kai Zhang, Deyu Meng, Radu Timofte, and Luc Van Gool. 2023. DDFM: Denoising Diffusion Model for Multi-Modality Image Fusion. InProceedings of the IEEE/CVF International Conference on Computer Vision...

    2023

  38. [46]

    Pengfei Zhu, Yang Sun, Bing Cao, and Qinghua Hu. 2024. Task-customized mixture of adapters for general image fusion. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 7099–7108. 10

    2024

Pith tools

Reviewed July 13, 2026 · model on record in the stance chip above.