OceanSplat: Object-aware Gaussian Splatting with Trinocular View Consistency for Underwater Scene Reconstruction

TL;DR

3D Gaussians are disentangled from the scattering medium, enabling robust representation of object geometry and significantly reducing floating artifacts in reconstructed underwater scenes and substantially outperforms existing methods for both scene reconstruction and restoration in scattering media.

Abstract

We introduce OceanSplat, a novel 3D Gaussian Splatting-based approach for accurately representing 3D geometry in underwater scenes. To overcome multi-view inconsistencies caused by underwater optical degradation, our method enforces trinocular view consistency by rendering horizontally and vertically translated camera views relative to each input view and aligning them via inverse warping. Furthermore, these translated camera views are used to derive a synthetic epipolar depth prior through triangulation, which serves as a self-supervised depth regularizer. These geometric constraints facilitate the spatial optimization of 3D Gaussians and preserve scene structure in underwater environments. We also propose a depth-aware alpha adjustment that modulates the opacity of 3D Gaussians during early training based on their $z$-component and viewing direction, deterring the formation of medium-induced primitives. With our contributions, 3D Gaussians are disentangled from the scattering medium, enabling robust representation of object geometry and significantly reducing floating artifacts in reconstructed underwater scenes. Experiments on real-world underwater and simulated scenes demonstrate that OceanSplat substantially outperforms existing methods for both scene reconstruction and restoration in scattering media.

Figures (5)

• Figure 1: OceanSplat overcomes scattering and attenuation effects through trinocular view consistency, preserving object geometry and enabling high-quality underwater 3D reconstruction.
• Figure 2: Overview of OceanSplat. We enforce trinocular view consistency by inverse warping rendered images from two translated camera poses to guide the spatial optimization of 3D Gaussians. From these poses, we derive a synthetic epipolar depth prior via triangulation, which provides self-supervised geometric constraints. Additionally, depth-aware alpha adjustment suppresses erroneous 3D Gaussians early and aligns rendered depth with the Gaussian $z$-component to prevent floating artifacts.
• Figure 3: Pixel-wise color contributions across translated camera views in 3D Gaussian Splatting.
• Figure 4: Qualitative evaluation of novel view synthesis on diverse real-world underwater 3D scenes.
• Figure 5: Qualitative comparison of scene reconstruction and restoration in underwater and foggy environments.