Efficient Burst Super-Resolution with One-step Diffusion

2025/07/18 by Kawai, Kento, Oba, Takeru, Tokoro, Kyotaro +2
#Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences

paper · doi:10.48550/arxiv.2507.13607

Abstract

While burst Low-Resolution (LR) images are useful for improving their Super Resolution (SR) image compared to a single LR image, prior burst SR methods are trained in a deterministic manner, which produces a blurry SR image. Since such blurry images are perceptually degraded, we aim to reconstruct sharp and high-fidelity SR images by a diffusion model. Our method improves the efficiency of the diffusion model with a stochastic sampler with a high-order ODE as well as one-step diffusion using knowledge distillation. Our experimental results demonstrate that our method can reduce the runtime to 1.6 % of its baseline while maintaining the SR quality measured based on image distortion and perceptual quality.

Citations

• Burst Super-Resolution with Diffusion Models for Improving Perceptual Quality
• Efficient Diffusion Model for Image Restoration by Residual Shifting
• SinSR: Diffusion-Based Image Super-Resolution in a Single Step
• Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusion
• ResShift: Efficient Diffusion Model for Image Super-resolution by Residual Shifting
• Exploiting Diffusion Prior for Real-World Image Super-Resolution
• Burstormer: Burst Image Restoration and Enhancement Transformer
• Consistency Models
• GibbsDDRM: A Partially Collapsed Gibbs Sampler for Solving Blind Inverse Problems with Denoising Diffusion Restoration
• Diffusion Model Based Posterior Sampling for Noisy Linear Inverse Problems
• Parallel Diffusion Models of Operator and Image for Blind Inverse Problems
• Diffusion Posterior Sampling for General Noisy Inverse Problems
• Improving Diffusion Models for Inverse Problems using Manifold Constraints
• Elucidating the Design Space of Diffusion-Based Generative Models
• BSRT: Improving Burst Super-Resolution with Swin Transformer and Flow-Guided Deformable Alignment
• Perception Prioritized Training of Diffusion Models
• Denoising Diffusion Restoration Models
• High-Resolution Image Synthesis with Latent Diffusion Models
• High-Resolution Image Synthesis with Latent Diffusion Models
• Come-Closer-Diffuse-Faster: Accelerating Conditional Diffusion Models\n for Inverse Problems through Stochastic Contraction
• Burst Image Restoration and Enhancement
• SDEdit: Guided Image Synthesis and Editing with Stochastic Differential Equations
• Video Swin Transformer
• Video Swin Transformer
• NTIRE 2021 Challenge on Burst Super-Resolution: Methods and Results
• Image Super-Resolution via Iterative Refinement
• Improved Denoising Diffusion Probabilistic Models
• Deep Burst Super-Resolution
• Score-Based Generative Modeling through Stochastic Differential Equations
• Denoising Diffusion Implicit Models
• AIM 2020 Challenge on Video Extreme Super-Resolution: Methods and Results
• Denoising Diffusion Probabilistic Models
• Space-Time-Aware Multi-Resolution Video Enhancement
• Deep Back-Projection Networks for Single Image Super-resolution
• Recurrent Back-Projection Network for Video Super-Resolution
• Recovering Realistic Texture in Image Super-resolution by Deep Spatial Feature Transform
• Deep Back-Projection Networks For Super-Resolution
• The Unreasonable Effectiveness of Deep Features as a Perceptual Metric
• Deformable Convolutional Networks
• Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network
• Image Super-Resolution Using Deep Convolutional Networks
• Image Super-Resolution Using Deep Convolutional Networks
• Image quality assessment: from error visibility to structural similarity
• Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
• Swin Transformer: Hierarchical Vision Transformer using Shifted Windows

Related

• Burst Super-Resolution with Diffusion Models for Improving Perceptual Quality
• ResShift: Efficient Diffusion Model for Image Super-resolution by Residual Shifting
• Arbitrary-steps Image Super-resolution via Diffusion Inversion
• SinSR: Diffusion-Based Image Super-Resolution in a Single Step
• FiDeSR: High-Fidelity and Detail-Preserving One-Step Diffusion Super-Resolution