Learning Hazing to Dehazing: Towards Realistic Haze Generation for Real-World Image Dehazing

2025-03-25

AI Summary

Existing image dehazing methods are constrained by the scarcity of realistic paired training data and the low sampling efficiency of diffusion models, particularly struggling to recover severely distorted details under dense fog. To address these challenges, this paper proposes a haze-generation–dehazing co-design framework. First, we introduce HazeGen—a realistic haze-image generator built upon a pretrained text-to-image diffusion model—to alleviate the data scarcity bottleneck. Second, we design a block-wise statistical alignment operator (AlignOp) and an acceleration-aware fidelity-preserving sampling mechanism (AccSamp), drastically reducing dehazing sampling steps without compromising reconstruction quality. By integrating hybrid training and fused sampling strategies, our method achieves state-of-the-art performance across multiple synthetic and real-world benchmarks, yielding more natural visual results and accelerating inference by 3–5×. The source code is publicly available.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Deep Generative Models & Autoencoders

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Web Mining and Content Analysis: Large pretrained models with web data
• Security and Privacy: Data transparency and provenance

Abstract

Existing real-world image dehazing methods primarily attempt to fine-tune pre-trained models or adapt their inference procedures, thus heavily relying on the pre-trained models and associated training data. Moreover, restoring heavily distorted information under dense haze requires generative diffusion models, whose potential in dehazing remains underutilized partly due to their lengthy sampling processes. To address these limitations, we introduce a novel hazing-dehazing pipeline consisting of a Realistic Hazy Image Generation framework (HazeGen) and a Diffusion-based Dehazing framework (DiffDehaze). Specifically, HazeGen harnesses robust generative diffusion priors of real-world hazy images embedded in a pre-trained text-to-image diffusion model. By employing specialized hybrid training and blended sampling strategies, HazeGen produces realistic and diverse hazy images as high-quality training data for DiffDehaze. To alleviate the inefficiency and fidelity concerns associated with diffusion-based methods, DiffDehaze adopts an Accelerated Fidelity-Preserving Sampling process (AccSamp). The core of AccSamp is the Tiled Statistical Alignment Operation (AlignOp), which can provide a clean and faithful dehazing estimate within a small fraction of sampling steps to reduce complexity and enable effective fidelity guidance. Extensive experiments demonstrate the superior dehazing performance and visual quality of our approach over existing methods. The code is available at https://github.com/ruiyi-w/Learning-Hazing-to-Dehazing.

Problem

• Generating realistic hazy images for dehazing training data
• Improving diffusion models for efficient dense haze removal
• Enhancing dehazing fidelity with accelerated sampling techniques

Innovation

• Realistic Hazy Image Generation framework
• Diffusion-based Dehazing framework
• Accelerated Fidelity-Preserving Sampling process