Source-Free Domain Adaptation for Real-world Image Dehazing

TL;DR

This paper introduces a source-free unsupervised domain adaptation method for real-world image dehazing, enabling existing models to adapt to real hazy images without source data, using a novel normalization module and unsupervised losses.

Contribution

The paper proposes a plug-and-play Domain Representation Normalization (DRN) module and unsupervised loss functions for source-free domain adaptation in image dehazing, addressing domain shift without source data.

Findings

Outperforms existing methods on multiple benchmarks.

Effectively bridges synthetic and real haze domain gaps.

Improves dehazing quality both visually and quantitatively.

Abstract

Deep learning-based source dehazing methods trained on synthetic datasets have achieved remarkable performance but suffer from dramatic performance degradation on real hazy images due to domain shift. Although certain Domain Adaptation (DA) dehazing methods have been presented, they inevitably require access to the source dataset to reduce the gap between the source synthetic and target real domains. To address these issues, we present a novel Source-Free Unsupervised Domain Adaptation (SFUDA) image dehazing paradigm, in which only a well-trained source model and an unlabeled target real hazy dataset are available. Specifically, we devise the Domain Representation Normalization (DRN) module to make the representation of real hazy domain features match that of the synthetic domain to bridge the gaps. With our plug-and-play DRN module, unlabeled real hazy images can adapt existing well-trained source networks. Besides, the unsupervised losses are applied to guide the learning of the DRN module, which consists of frequency losses and physical prior losses. Frequency losses provide structure and style constraints, while the prior loss explores the inherent statistic property of haze-free images. Equipped with our DRN module and unsupervised loss, existing source dehazing models are able to dehaze unlabeled real hazy images. Extensive experiments on multiple baselines demonstrate the validity and superiority of our method visually and quantitatively.