Learning Unpaired Image Dehazing with Physics-based Rehazy Generation

2025-06-15

AI Summary

Synthetic data overfitting severely degrades generalization to real-world hazy scenes in unsupervised image dehazing. Method: We propose a physics-driven unsupervised framework: (1) a rehazing strategy grounded in the atmospheric scattering model to generate high-fidelity hazy–rehazy paired data; and (2) a physically interpretable dual-branch progressive dehazing network that jointly models synthetic priors and real hazy image distributions. The architecture integrates coarse-to-fine feature restoration, an enhanced CycleGAN framework, and dual-branch collaborative training. Results: Our method achieves state-of-the-art performance among unsupervised approaches, improving PSNR by 3.58 dB on SOTS-Indoor and 1.85 dB on SOTS-Outdoor, significantly narrowing the domain gap between synthetic and real hazy images.

Technology Category

• Computer Vision: Computational Photography, Image & Video Synthesis
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Machine Learning: Unsupervised & Self-Supervised Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Overfitting to synthetic training pairs remains a critical challenge in image dehazing, leading to poor generalization capability to real-world scenarios. To address this issue, existing approaches utilize unpaired realistic data for training, employing CycleGAN or contrastive learning frameworks. Despite their progress, these methods often suffer from training instability, resulting in limited dehazing performance. In this paper, we propose a novel training strategy for unpaired image dehazing, termed Rehazy, to improve both dehazing performance and training stability. This strategy explores the consistency of the underlying clean images across hazy images and utilizes hazy-rehazy pairs for effective learning of real haze characteristics. To favorably construct hazy-rehazy pairs, we develop a physics-based rehazy generation pipeline, which is theoretically validated to reliably produce high-quality rehazy images. Additionally, leveraging the rehazy strategy, we introduce a dual-branch framework for dehazing network training, where a clean branch provides a basic dehazing capability in a synthetic manner, and a hazy branch enhances the generalization ability with hazy-rehazy pairs. Moreover, we design a new dehazing network within these branches to improve the efficiency, which progressively restores clean scenes from coarse to fine. Extensive experiments on four benchmarks demonstrate the superior performance of our approach, exceeding the previous state-of-the-art methods by 3.58 dB on the SOTS-Indoor dataset and by 1.85 dB on the SOTS-Outdoor dataset in PSNR. Our code will be publicly available.

Problem

• Overfitting to synthetic training pairs in image dehazing
• Training instability in unpaired dehazing methods
• Poor generalization to real-world hazy scenarios

Innovation

• Physics-based rehazy generation pipeline
• Dual-branch framework for dehazing
• Progressive coarse-to-fine dehazing network