Real-World Adverse Weather Image Restoration via Dual-Level Reinforcement Learning with High-Quality Cold Start

2025-11-07

AI Summary

Adverse weather severely degrades visual perception, and existing fixed-parameter models trained on synthetic data struggle to generalize to complex real-world degradation scenarios. To address this, we propose a two-level reinforcement learning–driven adaptive image restoration framework: an upper-level meta-controller dynamically selects and schedules restoration models, while a lower-level local optimizer performs no-reference image quality–guided perturbation optimization. Leveraging the physics-informed, high-fidelity synthetic dataset HFLS-Weather, our method enables cold-start initialization and supports online adaptation without paired supervision. Crucially, it eliminates reliance on real paired training data. Extensive experiments across diverse real-world adverse conditions—including rain, fog, snow, and haze—demonstrate significant performance gains over state-of-the-art methods, with strong generalization and continuous adaptability. The source code is publicly available.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Multiagent Systems: Adversarial Agents

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Adverse weather severely impairs real-world visual perception, while existing vision models trained on synthetic data with fixed parameters struggle to generalize to complex degradations. To address this, we first construct HFLS-Weather, a physics-driven, high-fidelity dataset that simulates diverse weather phenomena, and then design a dual-level reinforcement learning framework initialized with HFLS-Weather for cold-start training. Within this framework, at the local level, weather-specific restoration models are refined through perturbation-driven image quality optimization, enabling reward-based learning without paired supervision; at the global level, a meta-controller dynamically orchestrates model selection and execution order according to scene degradation. This framework enables continuous adaptation to real-world conditions and achieves state-of-the-art performance across a wide range of adverse weather scenarios. Code is available at https://github.com/xxclfy/AgentRL-Real-Weather

Problem

• Restoring real-world images degraded by diverse adverse weather conditions
• Overcoming limitations of synthetic-trained models with fixed parameters
• Enabling continuous adaptation to complex weather degradations without paired supervision

Innovation

• High-fidelity physics-driven dataset simulates weather phenomena
• Dual-level reinforcement learning optimizes restoration without paired supervision
• Meta-controller dynamically orchestrates model selection and execution order