UW-GS: Distractor-Aware 3D Gaussian Splatting for Enhanced Underwater Scene Reconstruction

TL;DR

This paper introduces UW-GS, a novel 3D Gaussian splatting method tailored for underwater scene reconstruction, addressing challenges like light scattering, absorption, and dynamic objects, and demonstrating superior performance on a new dataset.

Contribution

The paper presents UW-GS, a specialized Gaussian splatting approach for underwater scenes, incorporating color modeling, physics-based density control, and dynamic object handling, with a new dataset for validation.

Findings

Achieves up to 1.26dB PSNR improvement over existing methods.

Effectively models underwater light absorption and scattering effects.

Handles dynamic objects with a binary motion mask.

Abstract

3D Gaussian splatting (3DGS) offers the capability to achieve real-time high quality 3D scene rendering. However, 3DGS assumes that the scene is in a clear medium environment and struggles to generate satisfactory representations in underwater scenes, where light absorption and scattering are prevalent and moving objects are involved. To overcome these, we introduce a novel Gaussian Splatting-based method, UW-GS, designed specifically for underwater applications. It introduces a color appearance that models distance-dependent color variation, employs a new physics-based density control strategy to enhance clarity for distant objects, and uses a binary motion mask to handle dynamic content. Optimized with a well-designed loss function supporting for scattering media and strengthened by pseudo-depth maps, UW-GS outperforms existing methods with PSNR gains up to 1.26dB. To fully verify the effectiveness of the model, we also developed a new underwater dataset, S-UW, with dynamic object masks.