Underwater image enhancement via multiscale disentanglement strategy

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Feb. 19, 2025

Underwater images suffer from color casts, low illumination, and blurred details caused by light absorption scattering in water. Existing data-driven methods often overlook the scene characteristics of underwater imaging, limiting their expressive power. To address above issues, we propose a Multiscale Disentanglement Network (MD-Net) for Image Enhancement (UIE), which mainly consists radiance disentanglement (SRD) transmission map (TMD) modules. Specifically, MD-Net first disentangles original into three physical parameters are (clear image), map, global background light. The proposed network then reconstructs these images. Furthermore, introduces class adversarial learning between reconstructed to supervise accuracy network. Moreover, design multi-level fusion module (MFM) dual-layer weight estimation unit (DWEU) cast adjustment visibility enhancement. Finally, conduct extensive qualitative quantitative experiments on benchmark datasets, demonstrate that our approach outperforms other traditional state-of-the-art methods. Our code results available at: https://github.com/WYJGR/MD-Net .

Language: Английский

WEDM: Wavelet-Enhanced Diffusion with Multi-Stage Frequency Learning for Underwater Image Enhancement

Journal of Imaging, Journal Year: 2025, Volume and Issue: 11(4), P. 114 - 114

Published: April 9, 2025

Underwater image enhancement (UIE) is inherently challenging due to complex degradation effects such as light absorption and scattering, which result in color distortion a loss of fine details. Most existing methods focus on spatial-domain processing, often neglecting the frequency-domain characteristics that are crucial for effectively restoring textures edges. In this paper, we propose novel UIE framework, Wavelet-based Enhancement Diffusion Model (WEDM), integrates decomposition with diffusion models. The WEDM consists two main modules: Wavelet Color Compensation Module (WCCM) correction LAB space using discrete wavelet transform, (WDM), replaces traditional convolutions wavelet-based operations preserve multi-scale frequency features. By combining residual denoising frequency-specific reduces noise amplification high-frequency blurring. Ablation studies further demonstrate essential roles WCCM WDM improving fidelity texture Our framework offers robust solution underwater visual tasks, promising applications marine exploration ecological monitoring.

Language: Английский