Bridging Degradation Discrimination and Generation for Universal Image Restoration

2026-01-31

AI Summary

This work addresses the challenge of universal image restoration, which requires simultaneously handling diverse degradation types while recovering high-quality details—a key difficulty lying in adaptively integrating degradation information during the diffusion process to balance fidelity and perceptual quality. To this end, the authors propose the BDG method, which for the first time tightly couples fine-grained degradation discrimination with diffusion-based generation. Specifically, a multi-angle, multi-scale gray-level co-occurrence matrix (MAS-GLCM) is introduced to precisely characterize degradation type and severity. Furthermore, a novel three-stage diffusion training strategy—comprising generation, bridging, and restoration—is designed to dynamically fuse discriminative cues to guide the restoration process. Without modifying the model architecture, BDG achieves significant performance gains on both universal image restoration and real-world super-resolution benchmarks, effectively harmonizing fidelity and perceptual quality.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Universal image restoration is a critical task in low-level vision, requiring the model to remove various degradations from low-quality images to produce clean images with rich detail. The challenges lie in sampling the distribution of high-quality images and adjusting the outputs on the basis of the degradation. This paper presents a novel approach, Bridging Degradation discrimination and Generation (BDG), which aims to address these challenges concurrently. First, we propose the Multi-Angle and multi-Scale Gray Level Co-occurrence Matrix (MAS-GLCM) and demonstrate its effectiveness in performing fine-grained discrimination of degradation types and levels. Subsequently, we divide the diffusion training process into three distinct stages: generation, bridging, and restoration. The objective is to preserve the diffusion model's capability of restoring rich textures while simultaneously integrating the discriminative information from the MAS-GLCM into the restoration process. This enhances its proficiency in addressing multi-task and multi-degraded scenarios. Without changing the architecture, BDG achieves significant performance gains in all-in-one restoration and real-world super-resolution tasks, primarily evidenced by substantial improvements in fidelity without compromising perceptual quality. The code and pretrained models are provided in https://github.com/MILab-PKU/BDG.

Problem

• Universal Image Restoration
• Degradation Discrimination
• Image Degradation
• Low-level Vision
• Diffusion Model

Innovation

• Bridging Degradation discrimination and Generation
• MAS-GLCM
• diffusion model
• universal image restoration
• multi-degraded restoration