Retinex Meets Language: A Physics-Semantics-Guided Underwater Image Enhancement Network

TL;DR

This paper introduces PSG-UIENet, a novel underwater image enhancement network that combines physics-based illumination correction with language-informed semantic guidance, utilizing a new multimodal dataset and achieving superior results.

Contribution

The study presents the first integration of textual guidance and multimodal datasets into underwater image enhancement, improving perceptual quality and semantic consistency.

Findings

PSG-UIENet outperforms fifteen state-of-the-art methods on multiple datasets.

Constructed a large-scale image-text UIE dataset, LUIQD-TD, with 6,418 triplets.

Designed an Image-Text Semantic Similarity loss to enhance semantic consistency.

Abstract

Underwater images often suffer from severe degradation caused by light absorption and scattering, leading to color distortion, low contrast and reduced visibility. Existing Underwater Image Enhancement (UIE) methods can be divided into two categories, i.e., prior-based and learning-based methods. The former rely on rigid physical assumptions that limit the adaptability, while the latter often face data scarcity and weak generalization. To address these issues, we propose a Physics-Semantics-Guided Underwater Image Enhancement Network (PSG-UIENet), which couples the Retinex-grounded illumination correction with the language-informed guidance. This network comprises a Prior-Free Illumination Estimator and a Semantics-Guided Image Restorer. In particular, the restorer leverages the textual descriptions generated by the Contrastive Language-Image Pre-training (CLIP) model to inject high-level semantics for perceptually meaningful guidance. Since multimodal UIE data sets are not publicly available, we also construct a large-scale image-text UIE data set, namely, LUIQD-TD, which contains 6,418 image-reference-text triplets. To explicitly measure and optimize semantic consistency between textual descriptions and images, we further design an Image-Text Semantic Similarity (ITSS) loss function. To our knowledge, this study makes the first effort to introduce both textual guidance and the multimodal data set into UIE tasks. Extensive experiments on our data set and four publicly available data sets demonstrate that the proposed PSG-UIENet achieves superior or comparable performance against fifteen state-of-the-art methods.