Ref-DGS: Reflective Dual Gaussian Splatting

2026-03-08

AI Summary

Existing Gaussian splatting methods struggle to efficiently model near-field specular reflections, limiting the quality of reflective surface reconstruction and novel view synthesis. This work proposes a dual-Gaussian representation framework that decouples geometry reconstruction from reflectance modeling within a standard rasterization pipeline: near-field reflections are captured through the joint representation of geometric Gaussians and local reflective Gaussians, while a global environmental reflection field handles far-field contributions. To our knowledge, this is the first approach to achieve efficient and physically plausible joint modeling of both near- and far-field specular reflections without explicit ray tracing. Experiments demonstrate that our method significantly outperforms existing techniques on reflective scenes, delivering superior rendering quality and substantially faster training compared to ray-based Gaussian approaches.

Technology Category

• Computer Vision: Remote Sensing / Geospatial AI
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Reflective appearance, especially strong and typically near-field specular reflections, poses a fundamental challenge for accurate surface reconstruction and novel view synthesis. Existing Gaussian splatting methods either fail to model near-field specular reflections or rely on explicit ray tracing at substantial computational cost. We present Ref-DGS, a reflective dual Gaussian splatting framework that addresses this trade-off by decoupling surface reconstruction from specular reflection within an efficient rasterization-based pipeline. Ref-DGS introduces a dual Gaussian scene representation consisting of geometry Gaussians and complementary local reflection Gaussians that capture near-field specular interactions without explicit ray tracing, along with a global environment reflection field for modeling far-field specular reflections. To predict specular radiance, we further propose a lightweight, physically-aware adaptive mixing shader that fuses global and local reflection features. Experiments demonstrate that Ref-DGS achieves state-of-the-art performance on reflective scenes while training substantially faster than ray-based Gaussian methods.

Problem

• reflective appearance
• specular reflections
• surface reconstruction
• novel view synthesis
• Gaussian splatting

Innovation

• Reflective Rendering
• Gaussian Splatting
• Specular Reflection
• Rasterization
• Dual Representation