Image Restoration via Diffusion Models with Dynamic Resolution

2026-05-13

AI Summary

This work addresses the high computational cost of existing diffusion models in image restoration, noting that latent-space approaches often suffer from reduced efficiency due to repeated encoding and decoding. To overcome this limitation, we introduce a dynamic resolution mechanism into diffusion-based image restoration for the first time, dynamically projecting data onto low-dimensional subspaces to significantly accelerate inference. Building upon this insight, we propose SubDPS, SubDAPS, and their enhanced variant SubDAPS++, which effectively adapt the DPS and DAPS frameworks to this subspace paradigm. Extensive experiments across multiple image restoration tasks and datasets demonstrate that our methods achieve substantial gains in inference efficiency while maintaining or even improving reconstruction quality, outperforming current diffusion-based approaches in overall performance.

Abstract

Diffusion models (DMs) have exhibited remarkable efficacy in various image restoration tasks. However, existing approaches typically operate within the high-dimensional pixel space, resulting in high computational overhead. While methods based on latent DMs seek to alleviate this issue by utilizing the compressed latent space of a variational autoencoder, they require repeated encoder-decoder inference. This introduces significant additional computational burdens, often resulting in runtime performance that is even inferior to that of their pixel-space counterparts. To mitigate the computational inefficiency, this work proposes projecting data into lower-dimensional subspaces using dynamic resolution DMs to accelerate the inference process. We first fine-tune pre-trained DMs for dynamic resolution priors and adapt DPS and DAPS, which are two widely used pixel-space methods for general image restoration tasks, into the proposed framework, yielding methods we refer to as SubDPS and SubDAPS, respectively. Given the favorable inference speed and reconstruction fidelity of SubDAPS, we introduce an enhanced variant termed SubDAPS++ to further boost both reconstruction efficiency and quality. Empirical evaluations across diverse image datasets and various restoration tasks demonstrate that the proposed methods outperform recent DM-based approaches in the majority of experimental scenarios. The code is available at https://github.com/StarNextDay/SubDAPS.git.

Problem

• Image Restoration
• Diffusion Models
• Computational Efficiency
• Dynamic Resolution
• Latent Space

Innovation

• Dynamic Resolution
• Diffusion Models
• Image Restoration
• Latent Subspace
• Computational Efficiency