Pith. sign in

REVIEW 4 major objections 5 minor 93 references

Task-Generalized Adaptive Cross-Domain Learning for Multimodal Image Fusion

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read AdaSFFuse claims that replacing fixed wavelet filters with adaptive, learnable ones, coupled with a Mamba-style spatial-frequency fusion module, yields a single compact network that leads or near-leads four multimodal fusion benchmarks.

desk verdict Genuinely new Mamba+wavelet fusion backbone with plausible SOTA numbers; fix the undefined SSIM variant, missing error bars, and undefined loss term before trusting the details. read the letter →

arxiv 2508.15505 v1 pith:2DLS52YX submitted 2025-08-21 cs.CV

Mengyu Wang , Zhenyu Liu , Kun Li , Yu Wang , Yuwei Wang , Yanyan Wei , Fei Wang This is my paper
classification cs.CV
keywords multimodalimagefusionadaptivewavelettransformstatespacedualityMambafrequencydecouplinginfrared-visiblemulti-exposuremedical
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to establish that one task-general network, AdaSFFuse, can compete with and often beat task-specific state-of-the-art methods across four distinct image fusion problems: infrared-visible, multi-focus, multi-exposure, and medical fusion. Its bet is that the key to cross-task fusion is adaptive frequency handling: instead of relying on fixed wavelet or Fourier transforms, the network learns the decomposition itself, and instead of fusing only in the spatial domain, it fuses spatial and frequency features together through a 2D extension of Mamba's state-space-duality module. If correct, this makes high-quality multimodal fusion much cheaper—0.78M parameters and 82.81 GFLOPs—and removes the need to redesign the architecture for each fusion task.

What carries the argument

AdaWAT (Adaptive Approximate Wavelet Transform) is a learnable approximation of a 2D discrete wavelet transform built from grouped convolutions (low-pass) and dilated convolutions with task-chosen dilation rates (high-pass), producing LL/LH/HL/HH sub-bands that adapt to image content, with an inverse (AdaIWAT) via transpose convolution. The Spatial-Frequency Mamba Block is a fusion module whose core, AdaD-SSD, extends Mamba2's state-space duality to 2D: it feeds projected features through a spatial-aware branch (3x3 conv + SiLU) and a frequency-filtering branch (FFT, learned power-spectrum threshold, IFFT), then runs a 2D state update over intra- and inter-region interactions. The training l

What would settle it

Run the released AdaSFFuse code on a fixed subset of the LLVIP test split and recompute the eight metrics with the exact implementation the code ships; if the SSIM values still exceed 1.0 under a standard SSIM implementation, or if the reproduced MI/Qabf/SSIM numbers deviate from the paper's Table I by more than run-to-run noise, the state-of-the-art claim is not settled. Independently, check whether the Eq. 12 intensity target is implemented anywhere; if not, the stated training objective cannot be what produced the reported checkpoints.

Watch

Extended reading notes

Core claim

The central claim is that learnable frequency decoupling plus dual-domain fusion is sufficient to get state-of-the-art results across tasks. AdaWAT builds the low- and high-frequency filter bank from grouped convolutions and dilation-rate-modulated convolutions, so the wavelet bases adapt to each scene; AdaIWAT reconstructs via transpose convolution. The Spatial-Frequency Mamba Block runs two parallel branches—a 3x3-conv spatial branch and an FFT-based branch whose threshold on the power spectrum is learned—then splits the fused representation into the X, A, B, C components of a 2D state-space-duality update that propagates state within and between spatial subregions. On the four benchmarks,

Load-bearing premise

The headline results assume the eight evaluation metrics are computed in a consistent, valid way for every method, with no error bars; the 'SSIM' column reports values above 1.0 that standard SSIM cannot produce (indicating a non-standard variant that is never described), and Eq. 12's intensity target M(Im1, Im2) is never defined.

Editorial extensions

If this is right

  • A single architecture, without task-specific redesign, reaches leading results on infrared-visible, multi-focus, multi-exposure, and medical fusion, suggesting generic fusion is tractable when the frequency decomposition itself is learned.
  • The 0.78M-parameter model and 82.81 GFLOPs (67.5% lower than SwinFusion's 254.92 G) mean learnable wavelet decoupling could replace heavy Transformer backbones in practical fusion systems.
  • Ablations (Tables III–V) attribute the reported gains specifically to AdaWAT and to the spatial/frequency branches of AdaD-SSD; removing either lowers the quantitative scores.
  • Fused outputs from AdaSFFuse improve object detection and semantic segmentation over single modalities on the M3FD and MFNet datasets, so the fusion quality transfers to high-level vision tasks.

Reading between the lines

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

  • The paper never defines M(Im1, Im2) in Eq. 12, so the intensity loss is underspecified; as written, the training objective is not fully reproducible, and the relative weighting of the three loss terms may not be what was actually used.
  • The 'SSIM' column reports values above 1.0 (e.g., 1.51, 1.78), which standard SSIM cannot produce; the cited metric [87] is a quality index for multi-exposure fusion, so the SSIM column likely measures a different quantity, and cross-method comparisons on that column must be checked for consistency.
  • Because the network fuses only the luminance channel and copies the visible image's chrominance back, gains on MEF and IVF color output may be dominated by luminance fusion rather than genuine cross-modal color reasoning; a testable extension would fuse chrominance or predict it from both modalities.
  • The same adaptive wavelet plus Mamba pairing could be tested on imperfectly aligned inputs or video fusion, where per-frame frequency adaptation might reduce motion artifacts; the paper only evaluates already-aligned image pairs.
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 / 5 minor

Summary. The paper proposes AdaSFFuse, a unified multimodal image fusion framework built from an adaptive approximate wavelet transform (AdaWAT) and spatial-frequency Mamba blocks. The method is evaluated on infrared-visible, multi-exposure, multi-focus, and medical image fusion, with quantitative comparisons against ten recent methods using eight standard metrics, plus ablations and downstream detection/segmentation experiments. The paper claims state-of-the-art or second-best results on most metrics across all four tasks while keeping the model compact (0.78 M parameters).

Significance. If the reported results are reliable, AdaSFFuse would be a competitive general-purpose fusion backbone with a favorable efficiency-accuracy trade-off. The paper's strengths include extensive comparisons across four tasks, systematic ablations isolating the proposed components (Tables III--V), and an explicit complexity analysis (Fig. 7). The promise of public code and the use of standard training protocols add to the reproducibility potential. However, the central quantitative claim depends on a non-standard and undefined SSIM variant, an undefined term in the training loss, and the absence of statistical error bars, so the significance is currently conditional on correcting these issues.

major comments (4)
  1. [§IV.A.2, Tables I and II] The SSIM metric as reported is not defined and takes impossible values for standard SSIM (e.g., 1.51 in IVF, 1.78 in MFF). Reference [87] is a multi-exposure fusion quality index, not the SSIM defined in Sec. III-D. Because the 'best/second-best' tallies depend on this column, the central SOTA claim is not verifiable until the exact variant (e.g., sum of per-input SSIM, MEF-SSIM, or another formulation) is specified and all tables are recomputed with a conventional, bounded SSIM or with the stated variant clearly defined.
  2. [Eq. (12), §III-D] The intensity loss Lint uses M(Im1,Im2) as an 'element-wise aggregation operation specific to the MMIF task,' but M(·) is never defined. This term is part of the total training objective and therefore essential for reproducibility. Please specify the aggregation (e.g., max, mean, weighted sum) for each of the four fusion tasks, since a single M is claimed to hold across tasks.
  3. [Tables I--VII, §IV] No error bars, confidence intervals, or significance tests are reported. Several margins over baselines are small (e.g., IVF SD: 47.60 vs. 47.27; MIF EN: 5.07 vs. 5.08 for CDDFuse), so without run-to-run variance it is unclear whether the claimed improvements are meaningful. Please provide results over multiple seeds (at least three) or report standard deviations for the main tables.
  4. [§IV.G, Tables VI and VII] The downstream task experiments compare fusion output only against the original visible and infrared inputs; no other fusion methods are compared. This makes it impossible to conclude that AdaSFFuse is superior to existing fusion methods for detection or segmentation. Please include at least one or two SOTA fusion baselines (e.g., CDDFuse, SwinFusion) under the same downstream protocol.
minor comments (5)
  1. [§IV.C.3] The section is titled 'Multi-focus Image Fusion (MIF)' but should read 'MFF'; MIF is used for medical image fusion elsewhere. Please correct.
  2. [§IV.D, Fig. 6] The text refers to 'Him1 and Him1'; the second should presumably be 'Him2'.
  3. [Eq. (6)] The notation 'IFFT(Λ ⊙ ∥Λ∥2 > λ)' is ambiguous: it is unclear whether the thresholding is applied to the magnitude spectrum, the complex spectrum, or the reconstructed features. Please clarify with a precise formula.
  4. [§III-A] There is a typo: 'are fused by specific frequency-segmented fusion to from Hifuse' should be 'to form'.
  5. [Fig. 2] The architecture diagram is very dense and the legend (W, T, rDConv) is hard to read. Consider enlarging the figure and/or separating the AdaWAT and Mamba-block schematics.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: AdaSFFuse's SOTA claims rest on external test-set comparisons with standard losses; self-citations are architectural inspirations only.

full rationale

The paper's derivation chain is not circular. AdaSFFuse is an architecture-plus-loss proposal; its central quantitative claims are out-of-sample evaluations on public benchmarks (LLVIP, SICE/MEFB, Real-MFF/MFI-WHU/Lytro/MFFW, Harvard) against ten published baselines, with losses in Eqs. 10-13 (SSIM, Sobel-gradient texture, intensity) that are standard for fusion. The network parameters are trained on training splits and then evaluated on held-out test splits, so the reported metrics are not fitted variables or definitions in disguise. The only self-citations ([29], [70]) are prior uses of dilated convolutions / temporal-frequency SSD and are used as component inspiration, not as a uniqueness theorem or as proof of the SOTA claim; removing them would not change the empirical comparison. The undefined M(Im1,Im2) in Eq. 12 and the >1 SSIM values in Tables I-II are reporting/definition gaps that create correctness risk, but they are not circular: the paper does not define M in terms of the fused output or derive the tables from the loss. The Discussion section even concedes cases where fusion does not help downstream tasks, further confirming the evaluation is not constructed to force a favorable conclusion. Therefore no load-bearing circular step was found.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The paper's contribution is a deep network, so the ledger is a list of hyperparameters chosen by hand plus standard mathematical assumptions. The main gaps are the undefined M(·) and the unstated per-task tuning; no new physical entities are introduced.

free parameters (4)
  • loss weights mu1, mu2, mu3 = 10, 20, 20
    Chosen by hand in Eq. 13; no search or derivation is reported, and they control the balance between SSIM, texture, and intensity losses.
  • network depth C = 64
    Set in Sec IV.B without justification; controls feature width.
  • block counts N1, N2 = 2 and 4
    Depth of shallow and deep Spatial-Frequency Mamba stacks; chosen by hand.
  • dilation rates for high/low frequency enhancement = 1 and 3
    rDConv=3 for low-frequency and rDConv=1 for high-frequency in AdaWAT, set ad hoc without ablation.
assumptions (5)
  • standard math Wavelet decomposition in Eqs. 1-3 is a valid representation for 2D images via separable convolution kernels.
    Invoked to build AdaWAT in Sec III.B; relies on standard wavelet theory.
  • standard math FFT/IFFT and pointwise multiplication in the frequency domain (Eq. 6) implement a valid linear filter.
    Used in the frequency filtering branch T(·).
  • domain assumption Input images are grayscale and aligned; at test time, only the Y channel of RGB is fused and the Cb/Cr channels are copied from the visible image.
    Stated in Sec IV.B; this is a task-specific processing choice that may not hold for misaligned or color-critical inputs.
  • domain assumption The intensity loss target M(Im1, Im2) in Eq. 12 is a known task-specific aggregation, but it is never defined in the paper.
    The loss function relies on an unspecified operation, so the training objective is incomplete as stated.
  • ad hoc to paper The same loss weighting (mu1=10, mu2=mu3=20) and architecture are appropriate across IVF, MEF, MFF, and MIF.
    The paper does not show a hyperparameter search or per-task tuning; this is a load-bearing assumption for the 'task-generalized' claim.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Task-Generalized Adaptive Cross-Domain Learning for Multimodal Image Fusion." pith.science (2026). https://pith.science/paper/2DLS52YX

@misc{pith2026250815505,
  author       = {Pith},
  title        = {Pith review of: Task-Generalized Adaptive Cross-Domain Learning for Multimodal Image Fusion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2DLS52YX}},
  note         = {Machine review of arXiv:2508.15505}
}
read the original abstract

Multimodal Image Fusion (MMIF) aims to integrate complementary information from different imaging modalities to overcome the limitations of individual sensors. It enhances image quality and facilitates downstream applications such as remote sensing, medical diagnostics, and robotics. Despite significant advancements, current MMIF methods still face challenges such as modality misalignment, high-frequency detail destruction, and task-specific limitations. To address these challenges, we propose AdaSFFuse, a novel framework for task-generalized MMIF through adaptive cross-domain co-fusion learning. AdaSFFuse introduces two key innovations: the Adaptive Approximate Wavelet Transform (AdaWAT) for frequency decoupling, and the Spatial-Frequency Mamba Blocks for efficient multimodal fusion. AdaWAT adaptively separates the high- and low-frequency components of multimodal images from different scenes, enabling fine-grained extraction and alignment of distinct frequency characteristics for each modality. The Spatial-Frequency Mamba Blocks facilitate cross-domain fusion in both spatial and frequency domains, enhancing this process. These blocks dynamically adjust through learnable mappings to ensure robust fusion across diverse modalities. By combining these components, AdaSFFuse improves the alignment and integration of multimodal features, reduces frequency loss, and preserves critical details. Extensive experiments on four MMIF tasks -- Infrared-Visible Image Fusion (IVF), Multi-Focus Image Fusion (MFF), Multi-Exposure Image Fusion (MEF), and Medical Image Fusion (MIF) -- demonstrate AdaSFFuse's superior fusion performance, ensuring both low computational cost and a compact network, offering a strong balance between performance and efficiency. The code will be publicly available at https://github.com/Zhen-yu-Liu/AdaSFFuse.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

93 extracted references · 69 canonical work pages

  1. [87]

    Perceptual quality assessment for multi-exposure image fusion,

    K. Ma, K. Zeng, and Z. Wang, “Perceptual quality assessment for multi-exposure image fusion,” IEEE Transactions on Image Processing, vol. 24, no. 11, pp. 3345–3356, 2015

    work page 2015

  2. [1]

    Cddfuse: Correlation-driven dual-branch feature decomposition for multi-modality image fusion,

    Z. Zhao, H. Bai, J. Zhang, Y . Zhang, S. Xu, Z. Lin, R. Timofte, and L. Van Gool, “Cddfuse: Correlation-driven dual-branch feature decomposition for multi-modality image fusion,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 5906–5916

    2023

  3. [2]

    Equivariant multi-modality image fusion,

    Z. Zhao, H. Bai, J. Zhang, Y . Zhang, K. Zhang, S. Xu, D. Chen, R. Timofte, and L. Van Gool, “Equivariant multi-modality image fusion,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, June 2024

    2024

  4. [3]

    Swinfusion: Cross-domain long-range learning for general image fusion via swin transformer,

    J. Ma, L. Tang, F. Fan, J. Huang, X. Mei, and Y . Ma, “Swinfusion: Cross-domain long-range learning for general image fusion via swin transformer,” IEEE/CAA Journal of Automatica Sinica, vol. 9, no. 7, pp. 1200–1217, 2022

    2022

  5. [4]

    Infrared and visible image fusion based on deep decomposition network and saliency analysis,

    L. Jian, R. Rayhana, L. Ma, S. Wu, Z. Liu, and H. Jiang, “Infrared and visible image fusion based on deep decomposition network and saliency analysis,” IEEE Transactions on Multimedia , vol. 24, pp. 3314–3326, 2021

    2021

  6. [5]

    Different input resolutions and arbitrary output resolution: A meta learning-based deep framework for infrared and visible image fusion,

    H. Li, Y . Cen, Y . Liu, X. Chen, and Z. Yu, “Different input resolutions and arbitrary output resolution: A meta learning-based deep framework for infrared and visible image fusion,” IEEE Transactions on Image Processing, vol. 30, pp. 4070–4083, 2021

    2021

  7. [6]

    Lrrnet: A novel representation learning guided fusion network for infrared and visible images,

    H. Li, T. Xu, X.-J. Wu, J. Lu, and J. Kittler, “Lrrnet: A novel representation learning guided fusion network for infrared and visible images,” IEEE Transactions on Pattern Analysis and Machine Intelli- gence, vol. 45, no. 9, pp. 11 040–11 052, 2023

    2023

  8. [7]

    Dmef: Multi-exposure image fusion based on a novel deep decomposition method,

    K. Wu, J. Chen, and J. Ma, “Dmef: Multi-exposure image fusion based on a novel deep decomposition method,” IEEE Transactions on Multimedia, vol. 25, pp. 5690–5703, 2022

    2022

Show all 93 references
  1. [8]

    Mef-gan: Multi-exposure image fusion via generative adversarial networks,

    H. Xu, J. Ma, and X.-P. Zhang, “Mef-gan: Multi-exposure image fusion via generative adversarial networks,”IEEE Transactions on Image Processing, vol. 29, pp. 7203–7216, 2020

    2020

  2. [9]

    Multi-focus image fusion with a natural enhancement via a joint multi-level deeply supervised convolutional neural network,

    W. Zhao, D. Wang, and H. Lu, “Multi-focus image fusion with a natural enhancement via a joint multi-level deeply supervised convolutional neural network,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 29, no. 4, pp. 1102–1115, 2018

    2018

  3. [10]

    Mlnet: A multi-domain lightweight net- work for multi-focus image fusion,

    X. Nie, B. Hu, and X. Gao, “Mlnet: A multi-domain lightweight net- work for multi-focus image fusion,” IEEE Transactions on Multimedia , vol. 25, pp. 5565–5579, 2022

    2022

  4. [11]

    Matr: Multimodal medical image fusion via multiscale adaptive transformer,

    W. Tang, F. He, Y . Liu, and Y . Duan, “Matr: Multimodal medical image fusion via multiscale adaptive transformer,”IEEE Transactions on Image Processing, vol. 31, pp. 5134–5149, 2022

    2022

  5. [12]

    Msgfu- sion: Medical semantic guided two-branch network for multimodal brain image fusion,

    J. Wen, F. Qin, J. Du, M. Fang, X. Wei, C. P. Chen, and P. Li, “Msgfu- sion: Medical semantic guided two-branch network for multimodal brain image fusion,” IEEE Transactions on Multimedia, vol. 26, pp. 944–957, 2023

    2023

  6. [13]

    A general framework for image fusion based on multi-scale transform and sparse representation,

    Y . Liu, S. Liu, and Z. Wang, “A general framework for image fusion based on multi-scale transform and sparse representation,” Information fusion, vol. 24, pp. 147–164, 2015

    2015

  7. [14]

    Ddcgan: A dual-discriminator conditional generative adversarial network for multi- resolution image fusion,

    J. Ma, H. Xu, J. Jiang, X. Mei, and X.-P. Zhang, “Ddcgan: A dual-discriminator conditional generative adversarial network for multi- resolution image fusion,” IEEE Transactions on Image Processing , vol. 29, pp. 4980–4995, 2020

    2020

  8. [15]

    Target-aware dual adversarial learning and a multi-scenario multi- modality benchmark to fuse infrared and visible for object detection,

    J. Liu, X. Fan, Z. Huang, G. Wu, R. Liu, W. Zhong, and Z. Luo, “Target-aware dual adversarial learning and a multi-scenario multi- modality benchmark to fuse infrared and visible for object detection,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Re...

    2022

  9. [16]

    Low-light wheat image enhancement using an explicit inter- channel sparse transformer,

    Y . Wang, F. Wang, K. Li, X. Feng, W. Hou, L. Liu, L. Chen, Y . He, and Y . Wang, “Low-light wheat image enhancement using an explicit inter- channel sparse transformer,” Computers and Electronics in Agriculture , vol. 224, p. 109169, 2024

    2024

  10. [17]

    Vigt: proposal-free video grounding with a learnable token in the transformer,

    K. Li, D. Guo, and M. Wang, “Vigt: proposal-free video grounding with a learnable token in the transformer,” Science China Information Sciences, vol. 66, no. 10, p. 202102, 2023

    2023

  11. [18]

    Drop an octave: Reducing spatial redundancy in convo- lutional neural networks with octave convolution,

    Y . Chen, H. Fan, B. Xu, Z. Yan, Y . Kalantidis, M. Rohrbach, S. Yan, and J. Feng, “Drop an octave: Reducing spatial redundancy in convo- lutional neural networks with octave convolution,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2019, pp. 3435–3444

    2019

  12. [19]

    Watermarking digital 3-d volumes in the discrete fourier transform domain,

    V . Solachidis and I. Pitas, “Watermarking digital 3-d volumes in the discrete fourier transform domain,” IEEE Transactions on Multimedia , vol. 9, no. 7, pp. 1373–1383, 2007

    2007

  13. [20]

    Fourier transform-based scalable image quality measure,

    M. Narwaria, W. Lin, I. V . McLoughlin, S. Emmanuel, and L.-T. Chia, “Fourier transform-based scalable image quality measure,” IEEE Transactions on Image Processing, vol. 21, no. 8, pp. 3364–3377, 2012

    2012

  14. [21]

    Improving long-tailed pest classification using diffusion model-based data augmentation,

    M. Du, F. Wang, Y . Wang, K. Li, W. Hou, L. Liu, Y . He, and Y . Wang, “Improving long-tailed pest classification using diffusion model-based data augmentation,” Computers and Electronics in Agriculture, vol. 234, p. 110244, 2025

    2025

  15. [22]

    Image restoration via frequency selection,

    Y . Cui, W. Ren, X. Cao, and A. Knoll, “Image restoration via frequency selection,” IEEE Transactions on Pattern Analysis and Machine Intelli- gence, 2023

    2023

  16. [23]

    Spatial-frequency adaptive remote sensing image dehazing with mix- ture of experts,

    H. Shen, H. Ding, Y . Zhang, X. Cong, Z.-Q. Zhao, and X. Jiang, “Spatial-frequency adaptive remote sensing image dehazing with mix- ture of experts,” IEEE Transactions on Geoscience and Remote Sensing, 2024

    2024

  17. [24]

    A new framework for complex wavelet transforms,

    F. C. Fernandes, R. L. van Spaendonck, and C. S. Burrus, “A new framework for complex wavelet transforms,” IEEE Transactions on Signal Processing, vol. 51, no. 7, pp. 1825–1837, 2003

    2003

  18. [25]

    Medical image fusion based on convolutional neural networks and non-subsampled contourlet transform,

    Z. Wang, X. Li, H. Duan, Y . Su, X. Zhang, and X. Guan, “Medical image fusion based on convolutional neural networks and non-subsampled contourlet transform,” Expert Systems with Applications , vol. 171, p. 114574, 2021

    2021

  19. [26]

    Laplacian rede- composition for multimodal medical image fusion,

    X. Li, X. Guo, P. Han, X. Wang, H. Li, and T. Luo, “Laplacian rede- composition for multimodal medical image fusion,” IEEE Transactions on Instrumentation and Measurement , vol. 69, no. 9, pp. 6880–6890, 2020

    2020

  20. [27]

    Mdlatlrr: A novel decomposition method for infrared and visible image fusion,

    H. Li, X.-J. Wu, and J. Kittler, “Mdlatlrr: A novel decomposition method for infrared and visible image fusion,” IEEE Transactions on Image Processing, vol. 29, pp. 4733–4746, 2020

    2020

  21. [28]

    Selective frequency network for image restoration,

    Y . Cui, Y . Tao, Z. Bing, W. Ren, X. Gao, X. Cao, K. Huang, and A. Knoll, “Selective frequency network for image restoration,” in The Eleventh International Conference on Learning Representations , 2023

    2023

  22. [29]

    Frequency decoupling for motion magnification via multi-level isomorphic architecture,

    F. Wang, D. Guo, K. Li, Z. Zhong, and M. Wang, “Frequency decoupling for motion magnification via multi-level isomorphic architecture,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 18 984–18 994

    2024

  23. [30]

    Mambadfuse: A mamba- based dual-phase model for multi-modality image fusion,

    Z. Li, H. Pan, K. Zhang, Y . Wang, and F. Yu, “Mambadfuse: A mamba- based dual-phase model for multi-modality image fusion,”arXiv preprint arXiv:2404.08406, 2024

    2024 arXiv

  24. [31]

    Image fusion for the novelty rotating synthetic aperture system based on vision transformer,

    Y . Sun, X. Zhi, S. Jiang, G. Fan, X. Yan, and W. Zhang, “Image fusion for the novelty rotating synthetic aperture system based on vision transformer,” Information Fusion, vol. 104, p. 102163, 2024

    2024

  25. [32]

    Fusionmamba: Efficient remote sensing image fusion with state space model,

    S. Peng, X. Zhu, H. Deng, L.-J. Deng, and Z. Lei, “Fusionmamba: Efficient remote sensing image fusion with state space model,” IEEE Transactions on Geoscience and Remote Sensing , 2024

    2024

  26. [33]

    A task-guided, implicitly- searched and meta-initialized deep model for image fusion,

    R. Liu, Z. Liu, J. Liu, X. Fan, and Z. Luo, “A task-guided, implicitly- searched and meta-initialized deep model for image fusion,” IEEE JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 13 Transactions on Pattern Analysis and Machine Intelligence , vol. 46, no. 10, pp...

    2021

  27. [34]

    Searching a compact architecture for robust multi-exposure image fusion,

    Z. Liu, J. Liu, G. Wu, Z. Chen, X. Fan, and R. Liu, “Searching a compact architecture for robust multi-exposure image fusion,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 34, no. 7, pp. 6224– 6237, 2024

    2024

  28. [35]

    Multi-focus image fusion with a deep convolutional neural network,

    Y . Liu, X. Chen, H. Peng, and Z. Wang, “Multi-focus image fusion with a deep convolutional neural network,” Information Fusion, vol. 36, pp. 191–207, 2017

    2017

  29. [36]

    Deepfuse: A deep unsupervised approach for exposure fusion with extreme exposure image pairs,

    K. Ram Prabhakar, V . Sai Srikar, and R. Venkatesh Babu, “Deepfuse: A deep unsupervised approach for exposure fusion with extreme exposure image pairs,” in Proceedings of the IEEE international conference on computer vision, 2017, pp. 4714–4722

    2017

  30. [37]

    Ifcnn: A general image fusion framework based on convolutional neural network,

    Y . Zhang, Y . Liu, P. Sun, H. Yan, X. Zhao, and L. Zhang, “Ifcnn: A general image fusion framework based on convolutional neural network,” Information Fusion, vol. 54, pp. 99–118, 2020

    2020

  31. [38]

    An interactively reinforced paradigm for joint infrared-visible image fusion and saliency object detection,

    D. Wang, J. Liu, R. Liu, and X. Fan, “An interactively reinforced paradigm for joint infrared-visible image fusion and saliency object detection,” Information Fusion, vol. 98, p. 101828, 2023

    2023

  32. [39]

    Attention-guided global-local adversarial learning for detail-preserving multi-exposure image fusion,

    J. Liu, J. Shang, R. Liu, and X. Fan, “Attention-guided global-local adversarial learning for detail-preserving multi-exposure image fusion,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 32, no. 8, pp. 5026–5040, 2022

    2022

  33. [40]

    Learning a deep multi- scale feature ensemble and an edge-attention guidance for image fusion,

    J. Liu, X. Fan, J. Jiang, R. Liu, and Z. Luo, “Learning a deep multi- scale feature ensemble and an edge-attention guidance for image fusion,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 32, no. 1, pp. 105–119, 2021

    2021

  34. [41]

    Target- aware dual adversarial learning and a multi-scenario multi-modality benchmark to fuse infrared and visible for object detection,

    J. Liu, X. Fan, Z. Huang, G. Wu, R. Liu, W. Zhong, and Z. Luo, “Target- aware dual adversarial learning and a multi-scenario multi-modality benchmark to fuse infrared and visible for object detection,” in 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition , 20...

    2022

  35. [42]

    Improving misaligned multi-modality image fusion with one-stage progressive dense registra- tion,

    D. Wang, J. Liu, L. Ma, R. Liu, and X. Fan, “Improving misaligned multi-modality image fusion with one-stage progressive dense registra- tion,” IEEE Transactions on Circuits and Systems for Video Technology, 2024

    2024

  36. [43]

    Infrared and visible image fusion: From data compatibility to task adaption,

    J. Liu, G. Wu, Z. Liu, D. Wang, Z. Jiang, L. Ma, W. Zhong, and X. Fan, “Infrared and visible image fusion: From data compatibility to task adaption,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 47, no. 4, pp. 2349–2369, 2025

    2025

  37. [44]

    Mlfuse: Multi-scenario feature joint learning for multi- modality image fusion,

    J. Lei, J. Li, J. Liu, B. Wang, S. Zhou, Q. Zhang, X. Wei, and N. K. Kasabov, “Mlfuse: Multi-scenario feature joint learning for multi- modality image fusion,” IEEE Transactions on Multimedia , 2025

    2025

  38. [45]

    Robust one-stop multi- modality image registration-fusion-segmentation framework against mis- alignments and adversarial attacks,

    D. Wang, X. Jiao, J. Liu, and X. Fan, “Robust one-stop multi- modality image registration-fusion-segmentation framework against mis- alignments and adversarial attacks,” IEEE Transactions on Multimedia , 2025

    2025

  39. [46]

    Coconet: Coupled contrastive learning network with multi-level feature ensemble for multi- modality image fusion,

    J. Liu, R. Lin, G. Wu, R. Liu, Z. Luo, and X. Fan, “Coconet: Coupled contrastive learning network with multi-level feature ensemble for multi- modality image fusion,” International Journal of Computer Vision , vol. 132, no. 5, pp. 1748–1775, 2024

    2024

  40. [47]

    Deep rank- n decomposition network for image fusion,

    L. Sun, K. Zhang, F. Zhang, W. Wan, and J. Sun, “Deep rank- n decomposition network for image fusion,” IEEE Transactions on Multimedia, vol. 26, pp. 7335–7348, 2024

    2024

  41. [48]

    Ddfm: denoising diffusion model for multi-modality image fusion,

    Z. Zhao, H. Bai, Y . Zhu, J. Zhang, S. Xu, Y . Zhang, K. Zhang, D. Meng, R. Timofte, and L. Van Gool, “Ddfm: denoising diffusion model for multi-modality image fusion,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 8082–8093

    2023

  42. [49]

    Hitfusion: Infrared and visible image fusion for high-level vision tasks using transformer,

    J. Chen, J. Ding, and J. Ma, “Hitfusion: Infrared and visible image fusion for high-level vision tasks using transformer,” IEEE Transactions on Multimedia, 2024

    2024

  43. [50]

    Mufusion: A general unsupervised image fusion network based on memory unit,

    C. Cheng, T. Xu, and X.-J. Wu, “Mufusion: A general unsupervised image fusion network based on memory unit,” Information Fusion , vol. 92, pp. 80–92, 2023

    2023

  44. [51]

    Cgtf: Convolution-guided transformer for infrared and visible image fusion,

    J. Li, J. Zhu, C. Li, X. Chen, and B. Yang, “Cgtf: Convolution-guided transformer for infrared and visible image fusion,” IEEE Transactions on Instrumentation and Measurement , vol. 71, pp. 1–14, 2022

    2022

  45. [52]

    Itfuse: An interactive transformer for infrared and visible image fusion,

    W. Tang, F. He, and Y . Liu, “Itfuse: An interactive transformer for infrared and visible image fusion,” Pattern Recognition , vol. 156, p. 110822, 2024

    2024

  46. [53]

    Exploiting ensemble learning for cross-view isolated sign language recognition,

    F. Wang, K. Li, Y . Nie, Z. Duan, P. Zou, Z. Wu, Y . Wang, and Y . Wei, “Exploiting ensemble learning for cross-view isolated sign language recognition,” arXiv preprint arXiv:2502.02196 , 2025

    2025 arXiv

  47. [54]

    Probing synergistic high-order interaction in infrared and visible image fusion,

    N. Zheng, M. Zhou, J. Huang, J. Hou, H. Li, Y . Xu, and F. Zhao, “Probing synergistic high-order interaction in infrared and visible image fusion,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 26 384–26 395

    2024

  48. [55]

    Frequency- aware feature fusion for dense image prediction,

    L. Chen, Y . Fu, L. Gu, C. Yan, T. Harada, and G. Huang, “Frequency- aware feature fusion for dense image prediction,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 1, no. 1, pp. 1–18, 2024

    2024

  49. [56]

    Winnet: Wavelet-inspired invertible network for image denoising,

    J.-J. Huang and P. L. Dragotti, “Winnet: Wavelet-inspired invertible network for image denoising,” IEEE Transactions on Image Processing, vol. 31, pp. 4377–4392, 2022

    2022

  50. [57]

    Head-free lightweight semantic segmentation with linear transformer,

    B. Dong, P. Wang, and F. Wang, “Head-free lightweight semantic segmentation with linear transformer,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 37, no. 1, 2023, pp. 516–524

    2023

  51. [58]

    Spanet: Frequency- balancing token mixer using spectral pooling aggregation modulation,

    G. Yun, J. Yoo, K. Kim, J. Lee, and D. H. Kim, “Spanet: Frequency- balancing token mixer using spectral pooling aggregation modulation,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 6113–6124

    2023

  52. [59]

    Omni-frequency channel-selection representations for unsupervised anomaly detection,

    Y . Liang, J. Zhang, S. Zhao, R. Wu, Y . Liu, and S. Pan, “Omni-frequency channel-selection representations for unsupervised anomaly detection,” IEEE Transactions on Image Processing , 2023

    2023

  53. [60]

    Inception trans- former,

    C. Si, W. Yu, P. Zhou, Y . Zhou, X. Wang, and S. Yan, “Inception trans- former,” Advances in Neural Information Processing Systems , vol. 35, pp. 23 495–23 509, 2022

    2022

  54. [61]

    Mamba: Linear-time sequence modeling with selective state spaces,

    A. Gu and T. Dao, “Mamba: Linear-time sequence modeling with selective state spaces,” arXiv preprint arXiv:2312.00752 , 2023

    2023 arXiv

  55. [62]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” Advances in Neural Information Processing Systems , vol. 30, 2017

    2017

  56. [63]

    Eulermormer: Robust eulerian motion magnification via dynamic filtering within transformer,

    F. Wang, D. Guo, K. Li, and M. Wang, “Eulermormer: Robust eulerian motion magnification via dynamic filtering within transformer,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 38, no. 6, 2024, pp. 5345–5353

    2024

  57. [64]

    Repetitive action counting with hybrid temporal relation modeling,

    K. Li, X. Peng, D. Guo, X. Yang, and M. Wang, “Repetitive action counting with hybrid temporal relation modeling,” IEEE Transactions on Multimedia, 2025

    2025

  58. [65]

    Vision mamba: Efficient visual representation learning with bidirectional state space model,

    L. Zhu, B. Liao, Q. Zhang, X. Wang, W. Liu, and X. Wang, “Vision mamba: Efficient visual representation learning with bidirectional state space model,” arXiv preprint arXiv:2401.09417 , 2024

    2024 arXiv

  59. [66]

    Log-vmamba: Local- global vision mamba for medical image segmentation,

    T. D. Q. Dang, H. H. Nguyen, and A. Tiulpin, “Log-vmamba: Local- global vision mamba for medical image segmentation,” in Proceedings of the Asian Conference on Computer Vision , 2024, pp. 548–565

    2024

  60. [67]

    Transformers are SSMs: Generalized models and efficient algorithms through structured state space duality,

    T. Dao and A. Gu, “Transformers are SSMs: Generalized models and efficient algorithms through structured state space duality,” in Proceed- ings of the 41st International Conference on Machine Learning , vol. 235, 2024, pp. 10 041–10 071

    2024

  61. [68]

    Imagenet classification with deep convolutional neural networks,

    A. Krizhevsky, I. Sutskever, and G. E. Hinton, “Imagenet classification with deep convolutional neural networks,” Advances in Neural Informa- tion Processing Systems , vol. 25, 2012

    2012

  62. [69]

    Frequency-adaptive dilated con- volution for semantic segmentation,

    L. Chen, L. Gu, D. Zheng, and Y . Fu, “Frequency-adaptive dilated con- volution for semantic segmentation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 3414–3425

    2024

  63. [70]

    Temporal-frequency state space duality: An efficient paradigm for speech emotion recognition,

    J. Zhao, F. Wang, K. Li, Y . Wei, S. Tang, S. Zhao, and X. Sun, “Temporal-frequency state space duality: An efficient paradigm for speech emotion recognition,” in ICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing , 2025, pp. 1–5

    2025

  64. [71]

    Adaptive dynamic filtering network for image denoising,

    H. Shen, Z.-Q. Zhao, and W. Zhang, “Adaptive dynamic filtering network for image denoising,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 37, no. 2, 2023, pp. 2227–2235

    2023

  65. [72]

    Image quality assessment: from error visibility to structural similarity,

    Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli, “Image quality assessment: from error visibility to structural similarity,” IEEE transactions on image processing , vol. 13, no. 4, pp. 600–612, 2004

    2004

  66. [73]

    Visible and infrared image fusion using deep learning,

    X. Zhang and Y . Demiris, “Visible and infrared image fusion using deep learning,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 8, pp. 10 535–10 554, 2023

    2023

  67. [74]

    Learning a deep single image contrast enhancer from multi-exposure images,

    J. Cai, S. Gu, and L. Zhang, “Learning a deep single image contrast enhancer from multi-exposure images,” IEEE Transactions on Image Processing, vol. 27, no. 4, pp. 2049–2062, 2018

  68. [75]

    Benchmarking and comparing multi-exposure image fusion algorithms,

    X. Zhang, “Benchmarking and comparing multi-exposure image fusion algorithms,” Information Fusion, pp. 111–131, 2021

    2021

  69. [76]

    Real-mff: A large realistic multi-focus image dataset with ground truth,

    J. Zhang, Q. Liao, S. Liu, H. Ma, W. Yang, and J.-H. Xue, “Real-mff: A large realistic multi-focus image dataset with ground truth,” Pattern Recognition Letters, vol. 138, pp. 370–377, 2020

    2020

  70. [77]

    Mff-gan: An unsuper- vised generative adversarial network with adaptive and gradient joint constraints for multi-focus image fusion,

    H. Zhang, Z. Le, Z. Shao, H. Xu, and J. Ma, “Mff-gan: An unsuper- vised generative adversarial network with adaptive and gradient joint constraints for multi-focus image fusion,” Information Fusion, vol. 66, pp. 40–53, 2021

    2021

  71. [78]

    Multi-focus image fusion using dictionary-based sparse representation,

    M. Nejati, S. Samavi, and S. Shirani, “Multi-focus image fusion using dictionary-based sparse representation,” Information Fusion, vol. 25, pp. 72–84, 2015. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 14

    2015

  72. [79]

    Mffw: A new dataset for multi-focus image fusion,

    S. Xu, X. Wei, C. Zhang, J. Liu, and J. Zhang, “Mffw: A new dataset for multi-focus image fusion,” arXiv preprint arXiv:2002.04780 , 2020

    2002 arXiv

  73. [80]

    Assessment of image fusion procedures using entropy, image quality, and multispectral classification,

    J. W. Roberts, J. A. Van Aardt, and F. B. Ahmed, “Assessment of image fusion procedures using entropy, image quality, and multispectral classification,” Journal of Applied Remote Sensing , vol. 2, no. 1, p. 023522, 2008

    2008

  74. [81]

    In-fibre bragg grating sensors,

    Y .-J. Rao, “In-fibre bragg grating sensors,” Measurement Science and Technology, vol. 8, no. 4, p. 355, 1997

    1997

  75. [82]

    Image quality measures and their performance,

    A. M. Eskicioglu and P. S. Fisher, “Image quality measures and their performance,” IEEE Transactions on Communications , vol. 43, no. 12, pp. 2959–2965, 1995

    1995

  76. [83]

    A non- reference image fusion metric based on mutual information of image features,

    M. B. A. Haghighat, A. Aghagolzadeh, and H. Seyedarabi, “A non- reference image fusion metric based on mutual information of image features,” Computers & Electrical Engineering , vol. 37, no. 5, pp. 744– 756, 2011

    2011

  77. [84]

    Information measure for performance of image fusion,

    G. Qu, D. Zhang, and P. Yan, “Information measure for performance of image fusion,” Electronics letters, vol. 38, no. 7, p. 1, 2002

    2002

  78. [85]

    A new image fusion performance metric based on visual information fidelity,

    Y . Han, Y . Cai, Y . Cao, and X. Xu, “A new image fusion performance metric based on visual information fidelity,” Information fusion, vol. 14, no. 2, pp. 127–135, 2013

    2013

  79. [86]

    Objective image fusion performance measure,

    C. S. Xydeas, V . Petrovic et al. , “Objective image fusion performance measure,” Electronics letters, vol. 36, no. 4, pp. 308–309, 2000

    2000

  80. [88]

    Underwa- ter image enhancement using laplace decomposition,

    M. Iqbal, M. M. Riaz, S. S. Ali, A. Ghafoor, and A. Ahmad, “Underwa- ter image enhancement using laplace decomposition,” IEEE Geoscience and Remote Sensing Letters , vol. 19, pp. 1–5, 2020

    2020

  81. [89]

    YOLOv5 by Ultralytics,

    G. Jocher, “YOLOv5 by Ultralytics,” 2020. [Online]. Available: https://github.com/ultralytics/yolov5

    2020

  82. [90]

    Ultralytics YOLO,

    G. Jocher, J. Qiu, and A. Chaurasia, “Ultralytics YOLO,” 2023. [Online]. Available: https://github.com/ultralytics/ultralytics

    2023

  83. [91]

    Mfnet: Towards real-time semantic segmentation for autonomous vehicles with multi-spectral scenes,

    Q. Ha, K. Watanabe, T. Karasawa, Y . Ushiku, and T. Harada, “Mfnet: Towards real-time semantic segmentation for autonomous vehicles with multi-spectral scenes,” in 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . IEEE, 2017, pp. 5108–5115

    2017

  84. [92]

    Segformer: Simple and efficient design for semantic segmentation with transformers,

    E. Xie, W. Wang, Z. Yu, A. Anandkumar, J. M. Alvarez, and P. Luo, “Segformer: Simple and efficient design for semantic segmentation with transformers,” Advances in neural information processing systems , vol. 34, pp. 12 077–12 090, 2021

    2021

  85. [93]

    Hgformer: Hierarchical grouping transformer for domain generalized semantic seg- mentation,

    J. Ding, N. Xue, G.-S. Xia, B. Schiele, and D. Dai, “Hgformer: Hierarchical grouping transformer for domain generalized semantic seg- mentation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 15 413–15 423. Mengyu Wang receive...

    2023

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.