Sea-Undistort: A Dataset for Through-Water Image Restoration in High Resolution Airborne Bathymetric Mapping

2025-08-11

AI Summary

Shallow-water airborne bathymetric reconstruction faces severe optical distortions—including water surface waves, light scattering, and solar glare—while the scarcity of real paired training data hinders supervised learning. To address this, we introduce Sea-Undistort, the first high-resolution synthetic underwater image dataset tailored for airborne bathymetry, comprising 1,200 physically realistic clear–distorted image pairs with comprehensive metadata (water depth, solar azimuth, camera parameters), generated via physics-driven illumination and water-column modeling. Methodologically, we employ high-fidelity Blender rendering and propose a lightweight diffusion model incorporating a prior-guided glare mask for early-stage feature fusion. Evaluated on real aerial imagery, our approach significantly improves the completeness of digital seabed surface models, reduces bathymetric estimation error by 23.6%, effectively suppresses glare and scattering artifacts, and enables fine-grained recovery of sub-meter-scale benthic structures.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Distributed Search
• Intelligent Robots: State Estimation

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Accurate image-based bathymetric mapping in shallow waters remains challenging due to the complex optical distortions such as wave induced patterns, scattering and sunglint, introduced by the dynamic water surface, the water column properties, and solar illumination. In this work, we introduce Sea-Undistort, a comprehensive synthetic dataset of 1200 paired 512x512 through-water scenes rendered in Blender. Each pair comprises a distortion-free and a distorted view, featuring realistic water effects such as sun glint, waves, and scattering over diverse seabeds. Accompanied by per-image metadata such as camera parameters, sun position, and average depth, Sea-Undistort enables supervised training that is otherwise infeasible in real environments. We use Sea-Undistort to benchmark two state-of-the-art image restoration methods alongside an enhanced lightweight diffusion-based framework with an early-fusion sun-glint mask. When applied to real aerial data, the enhanced diffusion model delivers more complete Digital Surface Models (DSMs) of the seabed, especially in deeper areas, reduces bathymetric errors, suppresses glint and scattering, and crisply restores fine seabed details. Dataset, weights, and code are publicly available at https://www.magicbathy.eu/Sea-Undistort.html.

Problem

• Addressing optical distortions in underwater image restoration
• Creating synthetic dataset for supervised training in bathymetry
• Improving seabed mapping accuracy with enhanced diffusion models

Innovation

• Synthetic dataset for through-water image restoration
• Enhanced lightweight diffusion-based framework
• Early-fusion sun-glint mask for distortion reduction