OPERA: An Agent for Image Restoration with End-to-End Joint Planning-Execution Optimization

2026-05-21

AI Summary

Real-world image restoration is highly challenging due to the complex coupling of mixed degradations. Existing agent-based approaches are limited by narrow planning spaces and a lack of synergy among isolated, pre-trained restoration tools. This work proposes OPERA, a novel framework that achieves end-to-end joint optimization of planning and execution for the first time. OPERA employs reinforcement learning to directly optimize tool sequences within a combinatorial planning space and introduces an agent-guided co-training mechanism that enables restoration tools to learn collaborative behaviors when composed into sequences. By overcoming the bottlenecks of conventional methods, OPERA consistently outperforms current state-of-the-art universal models and agent-based approaches across multiple mixed-degradation benchmarks and real-world datasets, demonstrating superior and robust restoration performance.

Abstract

Real-world image restoration is challenging due to complex and interacting mixed degradations. Recent agent-based approaches address this problem by composing multiple task-specific restoration tools. However, empirical analysis reveals that their performance is fundamentally limited by implicitly constrained planning spaces and the lack of coordination among independently pretrained tools. To address these issues, we propose OPERA (Optimized Planning-Execution Restoration Agent), a framework that jointly optimizes restoration planning and tool execution in an end-to-end manner. On the planning side, OPERA uses reinforcement learning to directly optimize tool composition over a combinatorial plan space, with the final restoration quality as the reward. On the execution side, OPERA introduces agent-guided co-training of restoration tools, enabling them to learn cooperative behaviors under sequential composition. Extensive experiments on multi-degradation benchmarks and real-world datasets demonstrate that OPERA consistently outperforms both all-in-one restoration models and existing agent-based methods across diverse and complex degradation scenarios.

Problem

• image restoration
• mixed degradations
• agent-based approaches
• tool coordination
• planning space

Innovation

• end-to-end joint optimization
• reinforcement learning
• agent-guided co-training
• tool composition
• image restoration