DDTNet: Degradation Disentanglement and Transfer Network for Test-Time All-in-One De-weathering Adaptation

2026-06-15

AI Summary

Existing unified image deraining models exhibit performance trade-offs when handling diverse weather degradations—such as rain, fog, and snow—and suffer significant performance degradation under domain shift between training and testing conditions. To address this, this work proposes DDTNet, which shifts the focus from direct restoration of clean content to explicit modeling of degradation patterns. The core innovation lies in the Degradation Decoupling Module (DDM), which employs a Degradation-Coupled Attention (DCA) mechanism to jointly capture both generic and weather-specific features. This enables decoupling and transfer of target-domain degradation patterns and facilitates the generation of domain-adaptive paired data for fine-tuning. By establishing a test-time adaptive unified deraining framework, the proposed method consistently and substantially improves cross-domain performance of existing models across multiple real-world datasets.

Abstract

All-in-one adverse weather image restoration aims to remove multiple degradations, such as rain, haze, and snow, using a single unified model. Despite their broad applicability, existing methods typically compromise performance, delivering balanced but suboptimal results for individual degradation types. This issue becomes more pronounced when a domain gap exists between training and testing data. Motivated by the observation that modeling degradation patterns is more feasible than recovering clean content, we propose the Degradation Disentanglement and Transfer Network (DDTNet), which focuses specifically on degradation transfer. By disentangling degradation patterns from target-domain degraded images and transferring them to source domain clean images, DDTNet generates domain-adaptive paired training data. These pairs are then used to fine-tune restoration models, significantly enhancing their adaptability across diverse weather conditions and domains. The core of DDTNet is the Degradation Disentanglement Module (DDM), which comprises Degradation Coupled Attention (DCA) to capture both general and weather-specific features, thereby enabling effective disentanglement and transfer of degradation patterns. Experimental results demonstrate that DDTNet significantly and consistently improves existing all-in-one models across real-world deraining, desnowing, and dehazing datasets.

Problem

• all-in-one de-weathering
• domain gap
• image restoration
• degradation patterns
• test-time adaptation

Innovation

• Degradation Disentanglement
• Test-Time Adaptation
• All-in-One De-weathering
• Degradation Transfer
• Domain Adaptation