From Physical Degradation Models to Task-Aware All-in-One Image Restoration

TL;DR

This work addresses the challenge of performing multiple restoration tasks with a single model by grounding restoration in physical degradation and learning a task-aware inverse operator. It introduces OPIR, a two-stage framework that first predicts an inverse degradation operator and an uncertainty map, then refines restoration under uncertainty, all via a shared kernel-prediction backbone modulated by task embeddings. Key contributions include a task-aware module for per-task operator modulation, a principled uncertainty measure from kernel magnitudes, and a multi-scale, convolution-based operator implementation that maintains efficiency. The approach yields state-of-the-art all-in-one results and strong task-aligned performance across deraining, desnowing, and dehazing, with notable reductions in computational cost for real-time applicability.

Abstract

All-in-one image restoration aims to adaptively handle multiple restoration tasks with a single trained model. Although existing methods achieve promising results by introducing prompt information or leveraging large models, the added learning modules increase system complexity and hinder real-time applicability. In this paper, we adopt a physical degradation modeling perspective and predict a task-aware inverse degradation operator for efficient all-in-one image restoration. The framework consists of two stages. In the first stage, the predicted inverse operator produces an initial restored image together with an uncertainty perception map that highlights regions difficult to reconstruct, ensuring restoration reliability. In the second stage, the restoration is further refined under the guidance of this uncertainty map. The same inverse operator prediction network is used in both stages, with task-aware parameters introduced after operator prediction to adapt to different degradation tasks. Moreover, by accelerating the convolution of the inverse operator, the proposed method achieves efficient all-in-one image restoration. The resulting tightly integrated architecture, termed OPIR, is extensively validated through experiments, demonstrating superior all-in-one restoration performance while remaining highly competitive on task-aligned restoration.

Figures (7)

• Figure 1: Mechanisms of existing methods vs. our method.
• Figure 2: The figure shows t-SNE visualizations of degradation embeddings from OPIR (our method) and Defusion DefusionLuo_2025_CVPR, with our model displaying more distinct and well-separated clusters.
• Figure 3: The overall architecture of the proposed OPIR is composed of two stages.
• Figure 4: Qualitative results under the all-in-one experimental setup. Our OPIR recovers finer details in the reconstructed images.
• Figure 5: Qualitative results under the task-aligned experimental setup. Our OPIR is able to reconstruct finer and sharper details.