Underwater Diffusion Attention Network with Contrastive Language-Image Joint Learning for Underwater Image Enhancement

TL;DR

This work tackles the challenge of enhancing underwater images without reliable real paired references by proposing UDAN-CLIP, a diffusion-based image-to-image framework guided by vision-language priors. It pre-trains a diffusion model on a color-transfer synthetic underwater dataset (UIE-air) to retain natural in-air priors, then employs a CLIP-based, prompt-driven classifier and a spatial attention module to guide localized restoration and semantic consistency via a novel CLIP-Diffusion loss. The method combines domain-adaptive training, prompt learning for domain classification, and joint visual-textual alignment to mitigate artifacts and preserve details under challenging underwater conditions. Empirical results on multiple underwater benchmarks show that UDAN-CLIP achieves strong PSNR/SSIM and perceptual quality metrics, with qualitative visualizations demonstrating more natural color restoration and sharper structures, indicating improved applicability for downstream vision tasks.

Abstract

Underwater images are often affected by complex degradations such as light absorption, scattering, color casts, and artifacts, making enhancement critical for effective object detection, recognition, and scene understanding in aquatic environments. Existing methods, especially diffusion-based approaches, typically rely on synthetic paired datasets due to the scarcity of real underwater references, introducing bias and limiting generalization. Furthermore, fine-tuning these models can degrade learned priors, resulting in unrealistic enhancements due to domain shifts. To address these challenges, we propose UDAN-CLIP, an image-to-image diffusion framework pre-trained on synthetic underwater datasets and enhanced with a customized classifier based on vision-language model, a spatial attention module, and a novel CLIP-Diffusion loss. The classifier preserves natural in-air priors and semantically guides the diffusion process, while the spatial attention module focuses on correcting localized degradations such as haze and low contrast. The proposed CLIP-Diffusion loss further strengthens visual-textual alignment and helps maintain semantic consistency during enhancement. The proposed contributions empower our UDAN-CLIP model to perform more effective underwater image enhancement, producing results that are not only visually compelling but also more realistic and detail-preserving. These improvements are consistently validated through both quantitative metrics and qualitative visual comparisons, demonstrating the model's ability to correct distortions and restore natural appearance in challenging underwater conditions.

Figures (12)

• Figure 1: Architecture overview of the proposed UDAN-CLIP model that integrates vision-language alignment and a diffusion model to enhance an underwater image.
• Figure 2: Pre-training Diffusion Model with Prior Knowledge of Domain Adaption.
• Figure 3: Fine-tuning UDAN-CLIP with spatial attention and multi-classifier guidance strategy on real-underwater datasets. The intermediate results from $x_t \ \text{to} \ x_0$ move towards the in-air natural domain, mitigating the damage of fine-tuning to the prior knowledge of the pre-trained model.
• Figure 4: Performance metrics comparison of our proposed UDAN-CLIP model on the T200 dataset.
• Figure 5: Performance metrics comparison of our proposed UDAN-CLIP model on the Color-Checker7 dataset.