Underwater Image Enhancement via Dehazing and Color Restoration
Chengqin Wu, Shuai Yu, Tuyan Luo, Qiuhua Rao, Qingson Hu, Jingxiang Xu, Lijun Zhang
TL;DR
WaterFormer addresses the coupled yet independent problems of haze removal and color restoration in underwater imaging by leveraging a ViT-based three-branch architecture: DehazeFormer for haze features, CRB for color correction, and CFB for cross-feature fusion. It introduces a physics-guided soft reconstruction layer and two loss terms—Chromatic Consistency Loss and Sobel Color Loss—to preserve color fidelity and edge details under realistic water propagation, formalized with $U_ ext{λ}(x) = I_ ext{λ}(x) T_ ext{λ}(x) + A_ ext{λ} (1 - T_ ext{λ}(x))$ and $I_ ext{λ}(x) = U_ ext{λ}(x) K_ ext{λ}(x) - B_ ext{λ}(x) + U_ ext{λ}(x)$ where $K_ ext{λ}(x)= rac{1}{T_ ext{λ}(x)}-1$, $B_ ext{λ}(x) = A_ ext{λ} ( rac{1}{T_ ext{λ}(x)}-1)$. The method achieves superior full-reference metrics on synthetic and real datasets and demonstrates robust color fidelity and edge preservation, while ablations confirm the contribution of each component. The work advances underwater vision by decoupling degradation factors, guiding learning with physics-inspired priors, and offering practical gains for marine engineering tasks.
Abstract
Underwater visual imaging is crucial for marine engineering, but it suffers from low contrast, blurriness, and color degradation, which hinders downstream analysis. Existing underwater image enhancement methods often treat the haze and color cast as a unified degradation process, neglecting their inherent independence while overlooking their synergistic relationship. To overcome this limitation, we propose a Vision Transformer (ViT)-based network (referred to as WaterFormer) to improve underwater image quality. WaterFormer contains three major components: a dehazing block (DehazeFormer Block) to capture the self-correlated haze features and extract deep-level features, a Color Restoration Block (CRB) to capture self-correlated color cast features, and a Channel Fusion Block (CFB) that dynamically integrates these decoupled features to achieve comprehensive enhancement. To ensure authenticity, a soft reconstruction layer based on the underwater imaging physics model is included. Further, a Chromatic Consistency Loss and Sobel Color Loss are designed to respectively preserve color fidelity and enhance structural details during network training. Comprehensive experimental results demonstrate that WaterFormer outperforms other state-of-the-art methods in enhancing underwater images.