mlsanityData to 14 Oct 2026

Learning Where and What to Restore for Composite Image Restoration

arXiv 2503.17915 ↗First posted 23 Mar 2025eess.IV cs.AI cs.CVcited by 4

Not on the map. Why?

Abstract (arXiv)

Real-world degraded images often contain multiple co-occurring degradation types, making composite image restoration fundamentally different from the single-degradation setting assumed by most existing all-in-one methods. These methods typically apply uniform spatial computation and single-label task conditioning, limiting both spatial adaptivity and explicit modeling of degradation mixtures. We propose CART (Composite-Adaptive Routing and Task Conditioning), a unified framework that jointly decides where to spend computation and what degradation cues should guide restoration: a spatial mixer routes patches to graded-capacity experts by local restoration difficulty, and a channel mixer routes each image through degradation-specific experts, conditioned on the global task feature with multi-hot classification supervision. Empirically, CART achieves state-of-the-art performance for composite degradation restoration on CDD-11 and delivers the best results to date on the conventional 3-task and 5-task all-in-one benchmarks.

Key facts. Cited by 4 papers in the map (data through 14 Oct 2026). Builds on 37 of its 55 references that are in the map. First posted on arXiv 23 Mar 2025.

Builds on

37 of 55 references.

References · 37

  1. 2017Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layercited by 3,509
  2. 2021Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsitycited by 3,073
  3. 2016Categorical Reparameterization with Gumbel-Softmaxcited by 2,739
  4. 2013Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computationcited by 2,079
  5. 2021SwinIR: Image Restoration Using Swin Transformercited by 1,887
  6. 2021Restormer: Efficient Transformer for High-Resolution Image Restorationcited by 1,627
  7. 2021Swin Transformer V2: Scaling Up Capacity and Resolutioncited by 1,228
  8. 2020Pre-Trained Image Processing Transformercited by 893
  9. 2021Uformer: A General U-Shaped Transformer for Image Restorationcited by 866
  10. 2021Multi-Stage Progressive Image Restorationcited by 816
  11. 2016Deep Multi-scale Convolutional Neural Network for Dynamic Scene Deblurringcited by 727
  12. 2018Deep Retinex Decomposition for Low-Light Enhancementcited by 723
25 more
  1. 2020Prevalence of Neural Collapse during the terminal phase of deep learning trainingcited by 719
  2. 2022Simple Baselines for Image Restorationcited by 708
  3. 2017Benchmarking Single Image Dehazing and Beyondcited by 447
  4. 2016Deep Joint Rain Detection and Removal from a Single Imagecited by 352
  5. 2024MambaIR: A Simple Baseline for Image Restoration with State-Space Modelcited by 352
  6. 2020Learning Enriched Features for Real Image Restoration and Enhancementcited by 312
  7. 2021TransWeather: Transformer-based Restoration of Images Degraded by Adverse Weather Conditionscited by 244
  8. 2021A Geometric Analysis of Neural Collapse with Unconstrained Featurescited by 227
  9. 2022Restoring Vision in Adverse Weather Conditions with Patch-Based Denoising Diffusion Modelscited by 215
  10. 2020CycleISP: Real Image Restoration via Improved Data Synthesiscited by 213
  11. 2024InstructIR: High-Quality Image Restoration Following Human Instructionscited by 140
  12. 2022Deep Generalized Unfolding Networks for Image Restorationcited by 121
  13. 2024AdaIR: Adaptive All-in-One Image Restoration via Frequency Mining and Modulationcited by 111
  14. 2024OneRestore: A Universal Restoration Framework for Composite Degradationcited by 92
  15. 2024Complexity Experts are Task-Discriminative Learners for Any Image Restorationcited by 77
  16. 2023GridFormer: Residual Dense Transformer with Grid Structure for Image Restoration in Adverse Weather Conditionscited by 62
  17. 2023Multimodal Prompt Perceiver: Empower Adaptiveness, Generalizability and Fidelity for All-in-One Image Restorationcited by 50
  18. 2019Path-Restore: Learning Network Path Selection for Image Restorationcited by 40
  19. 2025Universal Image Restoration Pre-training via Degradation Classificationcited by 35
  20. 2024CAMixerSR: Only Details Need More "Attention"cited by 30
  21. 2024UniProcessor: A Text-induced Unified Low-level Image Processorcited by 22
  22. 2023Always Clear Days: Degradation Type and Severity Aware All-In-One Adverse Weather Removalcited by 12
  23. 2023DREAM: Diffusion Rectification and Estimation-Adaptive Modelscited by 12
  24. 2024Unified-Width Adaptive Dynamic Network for All-In-One Image Restorationcited by 6
  25. 2024AllRestorer: All-in-One Transformer for Image Restoration under Composite Degradationscited by 5

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 3; 2026: 1.

4 papers.

  1. 2025ClusIR: Towards Cluster-Guided All-in-One Image Restorationcited by 5
  2. 2025Clear Roads, Clear Vision: Advancements in Multi-Weather Restoration for Smart Transportationcited by 1
  3. 2025Pure-Pass: Fine-Grained, Adaptive Masking for Dynamic Token-Mixing Routing in Lightweight Image Super-Resolutioncited by 0
  4. 2026Beyond Uniform Restoration: Empowering All-in-One Restoration with Pixel-Level Multimodal Guidancecited by 0