Sea-ing Through Scattered Rays: Revisiting the Image Formation Model for Realistic Underwater Image Generation

2025-09-18

AI Summary

Existing underwater image generation models primarily address color fading, failing to model distance-dependent visibility degradation dominated by forward scattering in highly turbid conditions. This work proposes the first physically grounded generation framework that systematically integrates forward scattering modeling with heterogeneous medium representation. We further introduce BUCKET, the first controlled-turbidity, paired-annotation dataset. Our method extends the underwater imaging model by explicitly incorporating spatially varying scattering coefficients and depth-dependent attenuation, enabling fine-grained characterization of complex light transport in turbid media. Quantitative evaluation and user studies demonstrate that our generated images achieve significantly enhanced visual realism under high turbidity, with superior detail preservation. In pairwise preference tests, our method outperforms mainstream baselines with an 82.5% win rate, validating the critical importance of forward scattering modeling and heterogeneous medium representation for realistic underwater image synthesis.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Reasoning under Uncertainty: Uncertainty Representations
• Natural Language Processing: Generation

Application Category

• Web Mining and Content Analysis: Web data generation and simulation
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

In recent years, the underwater image formation model has found extensive use in the generation of synthetic underwater data. Although many approaches focus on scenes primarily affected by discoloration, they often overlook the model's ability to capture the complex, distance-dependent visibility loss present in highly turbid environments. In this work, we propose an improved synthetic data generation pipeline that includes the commonly omitted forward scattering term, while also considering a nonuniform medium. Additionally, we collected the BUCKET dataset under controlled turbidity conditions to acquire real turbid footage with the corresponding reference images. Our results demonstrate qualitative improvements over the reference model, particularly under increasing turbidity, with a selection rate of 82. 5% by survey participants. Data and code can be accessed on the project page: vap.aau.dk/sea-ing-through-scattered-rays.

Problem

• Improving underwater image formation model for turbid environments
• Including forward scattering and nonuniform medium effects
• Generating realistic synthetic data with visibility loss

Innovation

• Incorporating forward scattering term
• Considering nonuniform medium conditions
• Using controlled turbidity dataset collection