mlsanityData to 7 Oct 2026

ZipIR: Latent Pyramid Diffusion Transformer for High-Resolution Image Restoration

arXiv 2504.08591 ↗First posted 11 Apr 2025cs.CVcited by 5

On the map: Diffusion models → Inverse problems and image restoration

Abstract (arXiv)

Recent progress in generative models has significantly improved image restoration capabilities, particularly through powerful diffusion models that offer remarkable recovery of semantic details and local fidelity. However, deploying these models at ultra-high resolutions faces a critical trade-off between quality and efficiency due to the computational demands of long-range attention mechanisms. To address this, we introduce ZipIR, a novel framework that enhances efficiency, scalability, and long-range modeling for high-res image restoration. ZipIR employs a highly compressed latent representation that compresses image 32x, effectively reducing the number of spatial tokens, and enabling the use of high-capacity models like the Diffusion Transformer (DiT). Toward this goal, we propose a Latent Pyramid VAE (LP-VAE) design that structures the latent space into sub-bands to ease diffusion training. Trained on full images up to 2K resolution, ZipIR surpasses existing diffusion-based methods, offering unmatched speed and quality in restoring high-resolution images from severely degraded inputs.

Key facts. Cited by 5 papers in the map (data through 7 Oct 2026). Placed in Inverse problems and image restoration, one of 14 directions in Diffusion models. Builds on 52 of its 60 references that are in the map. First posted on arXiv 11 Apr 2025.

Builds on

52 of 60 references.

Closest ancestors · 4

Same direction, newest first

  1. 2023SeeSR: Towards Semantics-Aware Real-World Image Super-Resolution27 Nov
  2. 2023DiffBIR: Towards Blind Image Restoration with Generative Diffusion Prior29 Aug
  3. 2023ResShift: Efficient Diffusion Model for Image Super-resolution by Residual Shifting23 Jul
  4. 2023Exploiting Diffusion Prior for Real-World Image Super-Resolution11 May

Other references · 48

  1. 2017Attention Is All You Needcited by 59,932
  2. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  3. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 29,915
  4. 2020Denoising Diffusion Probabilistic Modelscited by 20,851
  5. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,694
  6. 2019Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformercited by 14,491
  7. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,029
  8. 2020Denoising Diffusion Implicit Modelscited by 9,655
  9. 2022Scalable Diffusion Models with Transformerscited by 5,931
  10. 2013Intriguing properties of neural networkscited by 5,114
  11. 2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesiscited by 4,290
  12. 2021Improved Denoising Diffusion Probabilistic Modelscited by 3,736
36 more
  1. 2020Taming Transformers for High-Resolution Image Synthesiscited by 3,229
  2. 2023InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Taskscited by 2,534
  3. 2017Non-local Neural Networkscited by 2,290
  4. 2021SwinIR: Image Restoration Using Swin Transformercited by 1,869
  5. 2023Consistency Modelscited by 1,778
  6. 2015DeepFool: a simple and accurate method to fool deep neural networkscited by 1,315
  7. 2018Demystifying MMD GANscited by 1,268
  8. 2014Image Super-Resolution Using Deep Convolutional Networkscited by 1,218
  9. 2021Cascaded Diffusion Models for High Fidelity Image Generationcited by 1,155
  10. 2021MUSIQ: Multi-scale Image Quality Transformercited by 1,058
  11. 2022Exploring CLIP for Assessing the Look and Feel of Imagescited by 974
  12. 2021Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Datacited by 961
  13. 2021Resolution-robust Large Mask Inpainting with Fourier Convolutionscited by 730
  14. 2024Improved Distribution Matching Distillation for Fast Image Synthesiscited by 648
  15. 2024Latte: Latent Diffusion Transformer for Video Generationcited by 536
  16. 2021Designing a Practical Degradation Model for Deep Blind Image Super-Resolutioncited by 524
  17. 2022MANIQA: Multi-dimension Attention Network for No-Reference Image Quality Assessmentcited by 514
  18. 2019Toward Real-World Single Image Super-Resolution: A New Benchmark and A New Modelcited by 400
  19. 2023InstaFlow: One Step is Enough for High-Quality Diffusion-Based Text-to-Image Generationcited by 387
  20. 2024Scaling Up to Excellence: Practicing Model Scaling for Photo-Realistic Image Restoration In the Wildcited by 281
  21. 2020Component Divide-and-Conquer for Real-World Image Super-Resolutioncited by 243
  22. 2023Retinexformer: One-stage Retinex-based Transformer for Low-light Image Enhancementcited by 228
  23. 2015Trainable Nonlinear Reaction Diffusion: A Flexible Framework for Fast and Effective Image Restorationcited by 194
  24. 2023Structural Pruning for Diffusion Modelscited by 193
  25. 2017PixelSNAIL: An Improved Autoregressive Generative Modelcited by 137
  26. 2021Mask-guided Spectral-wise Transformer for Efficient Hyperspectral Image Reconstructioncited by 110
  27. 2023DiffiT: Diffusion Vision Transformers for Image Generationcited by 104
  28. 2023Relay Diffusion: Unifying diffusion process across resolutions for image synthesiscited by 78
  29. 2022Any-resolution Training for High-resolution Image Synthesiscited by 72
  30. 2024LD-Pruner: Efficient Pruning of Latent Diffusion Models using Task-Agnostic Insightscited by 65
  31. 2024UltraPixel: Advancing Ultra-High-Resolution Image Synthesis to New Peakscited by 64
  32. 2024LinFusion: 1 GPU, 1 Minute, 16K Imagecited by 42
  33. 2024Diffusion-RWKV: Scaling RWKV-Like Architectures for Diffusion Modelscited by 41
  34. 2024MMCOMPOSITION: Revisiting the Compositionality of Pre-trained Vision-Language Modelscited by 29
  35. 2024FINECAPTION: Compositional Image Captioning Focusing on Wherever You Want at Any Granularitycited by 26
  36. 2022MANet: Improving Video Denoising with a Multi-Alignment Networkcited by 4

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 2; 2026: 3.

5 papers.

  1. 2025PixelDiT: Pixel Diffusion Transformers for Image Generationcited by 75
  2. 2025MMIG-Bench: Towards Comprehensive and Explainable Evaluation of Multi-Modal Image Generation Modelscited by 16
  3. 2026DuSPiT: Dual-Branch Sub-Patch Pixel Diffusion Transformercited by 1
  4. 2026LPM: Industrial-Scale Generative Video Restorationcited by 0
  5. 2026PixelDiT2: Representation-Grounded Pixel Diffusion Transformerscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Efficient One-Step Diffusion Restoration Model with Compact Token Compression and Linear Attention
  2. 2023Refusion: Enabling Large-Size Realistic Image Restoration with Latent-Space Diffusion Models
  3. 2025Harnessing Diffusion-Yielded Score Priors for Image Restoration
  4. 2025DiT4SR: Taming Diffusion Transformer for Real-World Image Super-Resolution
  5. 2026DiT-IC: Aligned Diffusion Transformer for Efficient Image Compression
  6. 2026PixRestore: Unified Image Restoration via Pixel Diffusion Transformer
  7. 2026When Latents Forget Pixels: Restoring Fidelity in Diffusion Transformer Super-Resolution
  8. 2023Image Super-resolution Via Latent Diffusion: A Sampling-space Mixture Of Experts And Frequency-augmented Decoder Approach