Eulerian Motion Reconstruction for Water Scenery

2026-09-29

AI Summary

This study addresses the challenge of reconstructing interactively renderable 3D dynamic water scenes from non-looping monocular videos. To this end, this work pioneers the construction of looping 4D dynamic scenes from a 3D perspective by proposing a reconstruction framework based on 3D Gaussian Splatting. Specifically, Gaussian primitives are advected through a static Eulerian motion field, while a time-varying residual term is introduced to accurately capture the authentic stochastic dynamics of water. The model is optimized under rendering loss supervision. This approach achieves high-fidelity reconstruction and novel-view rendering of 4D looping dynamic scenes from single-source videos. Both quantitative and qualitative evaluations demonstrate that the proposed method significantly outperforms existing techniques, yielding considerably more realistic dynamic water scene effects.

Abstract

Reconstructing and animating water scenery from nature produces compelling and immersive visual experiences. Previous work examined this task from the perspective of 2D video textures, with the goal of creating a looping video. In our work, we tackle the problem from a 3D perspective, creating a looping 4D dynamic reconstruction which can be interactively rendered from novel viewpoints from a single non-looping 2D source video. We represent motion as a 3D static \textit{Eulerian} motion field that advects canonical Gaussian splats that are cyclically reborn at fixed time periods, supervised using rendering losses. To model non-periodic and stochastic dynamics present in real-world scenes, we add a non-periodic, time-varying residual term to capture deviations from the static Eulerian motion field. We show quantitatively and qualitatively that our framework enables photorealistic animation of water scenes better than prior art.

Problem

• Water Scenery Reconstruction
• 4D Dynamic Reconstruction
• Eulerian Motion Field
• Novel View Synthesis
• Stochastic Dynamics

Innovation

• Eulerian Motion Field
• 4D Dynamic Reconstruction
• Gaussian Splats
• Water Scenery Animation
• Non-periodic Residual