Ocean4D: Generative Underwater 4D Reconstruction via Medium-Aware Video Diffusion

2026-06-22

AI Summary

Underwater 4D reconstruction is highly challenging due to light absorption, backscattering in participating media, and dynamic water flow disturbances. Existing approaches often rely on assumptions valid in air, compromising geometric stability and cross-view consistency. This work proposes Ocean4D—the first generative framework that integrates medium-aware mechanisms into a video diffusion model. Ocean4D enforces inter-frame structural coherence through 4D Geometric Consistency Constraints (4D-GCC) and implicitly models underwater optical degradation effects in latent space. Requiring only a monocular video and the target camera trajectory, it generates globally consistent and geometrically stable dynamic underwater scenes. Experiments demonstrate that Ocean4D significantly outperforms current methods on both static and dynamic underwater benchmarks, effectively handling floating particles and dynamic disturbances in turbid waters.

Abstract

Underwater 4D reconstruction remains challenging due to the coupling between degraded light transport in participating media and dynamic water variations. Most existing Methods are developed under in-air assumptions and do not explicitly account for underwater absorption and backscatter. Additionally, near-static assumptions make these approaches sensitive to drifting particles and dynamic distractors , leading to unstable geometry and inconsistent cross-view results. To address these issues, we propose a generative framework for underwater 4D reconstruction, named Ocean4D, which is built on two complementary components. Specifically, 4D-GCC constructs 4D geometrically consistent conditioning with improved cross-frame coverage, while the Medium-Aware Block performs implicit medium-aware denoising in the latent diffusion process to stabilize underwater appearance under absorption and scattering. Given a monocular video and target cameras, our method generates videos along the target trajectories while preserving global structure and cross-view consistency. Extensive experiments on both dynamic and static underwater benchmarks demonstrate state-of-the-art performance on underwater reconstruction.

Problem

• underwater 4D reconstruction
• light transport degradation
• absorption and backscatter
• dynamic distractors
• cross-view inconsistency

Innovation

• underwater 4D reconstruction
• medium-aware diffusion
• geometrically consistent conditioning
• latent video diffusion
• participating media modeling