Fusing Transferred Priors and Physics-based Decomposition for Underwater Image Enhancement

2026-06-14

AI Summary

This work addresses the challenges of underwater image degradation—such as color distortion, low contrast, and haze—caused by complex aquatic environments. Existing learning-based methods often rely on noisy pseudo-labels, yielding unreliable supervision. To overcome this limitation, the authors propose a novel unsupervised enhancement framework that decomposes the task according to a physical imaging model into three stages: global color correction, dehazing, and background noise suppression. Crucially, the method leverages cross-task transfer priors from related visual domains as unpaired supervisory signals. This approach uniquely integrates physics-driven task decomposition with cross-domain transfer priors, achieving state-of-the-art performance across multiple quantitative and qualitative benchmarks. Moreover, it significantly improves the efficacy of downstream vision tasks, demonstrating both theoretical soundness and strong generalization capability.

Abstract

The underwater images are captured within diverse water-medium conditions, leading to complex degradation, including color bias, low contrast, and blur effect. Recently, learning-based methods have demonstrated their potential for underwater image enhancement (UIE). However, most of the previous work focus on the training strategy or network design to make the enhanced result aligned well with the labels in datasets, ignoring that the labels are selected from the enhanced results of previous UIE methods and these pseudo-labels are noisy. Consequently, the performance of their models is not satisfactory to a certain extent. However, collecting the true labels of the underwater images is challenging. In this work, we propose a transfer learning-based UIE that does not require underwater images to have paired noisy or true labels for learning. Instead, the UIE task is first divided into global color correction, haze removal, and background noise suppression following the underwater physics. Then multiple types of prior from other vision tasks are leveraged as cross-domain supervision in each step. In this way, a novel UIE is available via transfer learning, and the physics-aligned UIE decomposition provides theoretical soundness. Qualitative and quantitative experiments demonstrate that our proposal based on physics and priors fusion achieves SOTA performance in the UIE task and effectively boosts downstream vision tasks, significantly outperforming benchmark methods. Project repo: https://github.com/Haru2022/P2-UIE.

Problem

• underwater image enhancement
• noisy pseudo-labels
• true label scarcity
• color bias
• low contrast

Innovation

• transfer learning
• physics-based decomposition
• cross-domain priors
• underwater image enhancement
• label-free learning