[论文解读] Bridging Degradation Discrimination and Generation for Universal Image Restoration

JiaKui Hu, Zhengjian Yao+2 more|arXiv (Cornell University)|Jan 31, 2026

Advanced Image Processing Techniques被引用 0

一句话总结

BDG 引入 MAS-GLCM 用于细粒度降解区分，并提出三阶段扩散训练框架，将降解感知与生成先验桥接，以实现普适图像修复。

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.

研究动机与目标

• 解决在普适图像修复中用单一模型处理多种降解的需求。
• 开发一个细粒度的降解表征方法，以区分降解类型和级别。
• 在不改变架构的前提下，将降解区分与基于扩散的生成结合起来。
• 在一体化与真实场景中，保留生成先验并提升修复保真度。
• 实现对混合降解和真实场景下的稳健修复。

提出的方法

• 引入 MAS-GLCM，即多角度多尺度灰度共生矩阵，用于区分降解类型和级别。
• 重构扩散的反向过程并实现三阶段训练：生成预训练、桥接、修复微调。
• 通过对齐 MAS-GLCM 特征与扩散特征，使用桥接损失和降解分类损失将降解区分与生成先验桥接起来。
• 在扩散模型中将残差和低质量输入作为条件，以实现有条件的修复。
• 使用一体化的 5D 修复设定进行训练，并在真实世界与混合降解任务上评估。
• 引入基于顺序的伪标签来表示降解进程，以模拟真实世界降解的复杂性。

Figure 1: (1) Visualization of MAS-GLCM in varying degradation levels. With an increase in degradation levels, the MAS-GLCM exhibits significant transformations. (2) The results of the T-SNE analysis for LQ images and MAS-GLCM across various degradation types demonstrate that MAS-GLCM possesses an e

实验结果

研究问题

• RQ1 MAS-GLCM 是否能提供比现有表征更细粒度的降解区分？
• RQ2 三阶段 BDG 扩散训练在保持生成先验的同时，是否能在多样降解场景中提升修复保真度？
• RQ3 在一体化、混合降解和真实世界超分任务中，BDG 相较于最先进方法的表现如何？
• RQ4 桥接与降解分类损失对修复质量与保真度有何影响？
• RQ5 真实世界降解顺序是否可用于训练稳健的修复模型？

主要发现

• BDG 在一体化修复任务（去雨、低照度增强、去雾与去模糊）上达到最先进水平。
• 在混合降解场景中，BDG 相较 DiffUIR 有显著提升，并在若干指标（如 PSNR/SSIM）上取得改进，尤以雾霾和降雨情形突出。
• 在真实世界 SR 任务中，BDG 在全参考指标（PSNR/SSIM/LPIPS）上获得一流或接近一流的结果，且在非参考指标（MANIQA、MUSIQ、CLIPIQA）上具有竞争力。
• 消融研究表明，桥接阶段和修复微调阶段对于达到最优性能是必需的，降解分类损失可防止桥接阶段的编码器塌陷。
• MAS-GLCM 相较 Sobel、Laplace 和傅里叶等特征，提供更优的降解类型与级别分类精度。
• 该模型在保持丰富纹理生成的同时，通过将降解区分与生成先验对齐，提升修复保真度。

Figure 2: Three training stages in BDG. (1) During the generation stage, the model focuses on obtaining generation priors. (2) In the bridging stage, the MAS-GLCM, which can identify degradation fine-grainedly, is aligned with the features of the pre-trained generation model, thereby endowing the mo