Lune

CVPR Findings2026

Blockwise Divide-and-Aggregate for Image Restoration using Diffusion Priors

Vishal Purohit, Wei Chen, Qiang Qiu

Abstract

Image restoration from degraded measurements requires recovering both measurement-consistent components and plausible structures lying in the null space of the forward operator. Due to the inherent ill-posedness of such problems, the underlying posterior distribution is often complex, admitting multiple plausible reconstructions consistent with the same observation. Leveraging diffusion models as powerful generative priors, sampling from different initial noise realizations yields diverse inference trajectories, each tracing a distinct mode of the posterior. To harness this diversity, we propose a sampler that initiates multiple generative trajectories from independent noise initializations, each exploring different possible reconstructions. To further enrich structural diversity and encourage exploration of varied spatial configurations, we introduce a patchwise divide-and-aggregate strategy, where the image is divided into local patches that are randomly rotated or permuted across trajectories. The locally restored patches are then aligned and fused through an aggregation step, enforcing global measurement fidelity. We validate our approach on diverse inverse problems -including inpainting, superresolution, compressed sensing, and Gaussian deblurringshowing consistent gains in reconstruction quality and uncertainty representation on LSUN-Bedroom and ImageNet benchmarks.

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

Questions to start from

Lune can draw on

the full paper · related work

Get the answer

Free to start. No credit card required.

Lineage

  • Palette: Image-to-Image Diffusion Models cites Diffusion Models Beat GANs on Image Synthesis
  • Palette: Image-to-Image Diffusion Models cites Denoising Diffusion Probabilistic Models
  • Diffusion Models Beat GANs on Image Synthesis cites Denoising Diffusion Probabilistic Models
  • Diffusion Models Beat GANs on Image Synthesis cites Denoising Diffusion Implicit Models
  • Denoising Diffusion Restoration Models cites Palette: Image-to-Image Diffusion Models
  • Denoising Diffusion Restoration Models cites Diffusion Models Beat GANs on Image Synthesis
  • Denoising Diffusion Restoration Models cites Denoising Diffusion Probabilistic Models
  • Denoising Diffusion Restoration Models cites Denoising Diffusion Implicit Models
  • Denoising Diffusion Implicit Models cites Denoising Diffusion Probabilistic Models
Builds onCited byLeft to right by year; the more cited, the higher.

Builds on22

  1. Denoising Diffusion Probabilistic ModelsJonathan Ho, Ajay Jain, Pieter AbbeelNeurIPS 2020 · 35,902 citations
  2. Diffusion Models Beat GANs on Image SynthesisPrafulla Dhariwal, Alexander Quinn NicholNeurIPS 2021 · 13,211 citations
  3. Denoising Diffusion Implicit ModelsJiaming Song, Chenlin Meng, Stefano ErmonICLR 2021 · 11,743 citations
  4. Palette: Image-to-Image Diffusion ModelsChitwan Saharia, William Chan, Huiwen Chang, Chris A. Lee et al.SIGGRAPH 2022 · 1,638 citations
  5. Denoising Diffusion Restoration ModelsBahjat Kawar, Michael Elad, Stefano Ermon, Jiaming SongNeurIPS 2022 · 1,439 citations

Related papers

  1. Kernel Density Steering: Inference-Time Scaling via Mode Seeking for Image RestorationYuyang Hu, Kangfu Mei, Mojtaba Sahraee-Ardakan, Ulugbek Kamilov et al.NeurIPS 2025 · 6 citations
  2. DGSolver: Diffusion Generalist Solver with Universal Posterior Sampling for Image RestorationHebaixu Wang, Jing Zhang, Haonan Guo, Di Wang et al.NeurIPS 2025 · 7 citations
  3. Improving Diffusion-Based Image Restoration with Error Contraction and Error CorrectionQiqi Bao, Zheng Hui, Rui Zhu, Peiran Ren et al.AAAI 2024 · 5 citations
  4. Consistency Posterior Sampling for Diverse Image SynthesisVishal Purohit, Matthew Repasky, Jianfeng Lu, Qiang Qiu et al.CVPR 2025
  5. Diffusion Posterior Sampling for General Noisy Inverse ProblemsHyungjin Chung, Jeongsol Kim, Michael Thompson McCann, Marc Louis Klasky et al.ICLR 2023 · 152 citations
Dusk over the sea between two cliffs drawn in fine vertical lines