RGS: Reflection-aware Gaussian Splatting via Learning Geometry Continuity for Reflective Objects

ICRA 2026



Xiaobiao Du     Yida Wang      Cheng Bi      Kun Zhan      Xin Yu

Motivation: Existing state-of-the-art reflective 3DGS approaches, 3DGS-DR in (d), struggle to precisely reconstruct the reflective surface, leading to unsatisfactory specular novel view synthesis. Traditional foundation models cannot estimate the correct normal in (a) and depth in (b) due to view-dependent ambiguity. We propose to use VGGT in (c) with cross-view geometry constraints for accurate reflective surface supervision. Thus, our proposed RGS in (e) is a simple yet effective method to regularize the specular surface for reflective objects. Our RGS achieves high-quality novel-view synthesis compared to its ground truth in (f).

Abstract

Gaussian Splatting has significantly improved the quality of novel view synthesis with explicit Gaussian representation. However, we observed that existing 3D Gaussian Splatting methods (3DGS) often suffer from surface collapse issues on reflective regions, and thus produce inferior geometry and low-quality specular. In this work, we propose a physically-based deferred rendering framework, named Reflection-aware Gaussian Splatting (RGS), that can accurately model specular regions and improve novel view synthesis performance. Specifically, we found that a powerful 3D foundation model can provide a strong 3D geometric prior to foster correct geometric modeling. Based on this, we propose a cross-view shape consistency regularization to regularize the geometry surface with the large model prior and cross-view constraints. In this manner, our RGS can produce smoother geometric surfaces on reflective regions while reducing geometric hollows. To further improve rendering results on reflective regions, we present a reflection-aware densification strategy that is designed to capture specular variations across various views. With this strategy, our RGS is able to render novel views of objects in higher quality. Extensive experiments demonstrate our method consistently renders high-quality reflective objects, achieving state-of-the-art performance.

Method

Pipeline of our proposed RGS. Our method with the 2D Gaussian representation can render appearance and geometry maps with a learnable HDR map. We utilize VGGT inherent with cross-view constraints to provide accurate geometry regularization. RGS typically focuses on the specular capture for high-quality novel view synthesis.

Novel View Synthesis Comparisons

Here, we provide the novel view synthesis comparisons of our proposed RGS with existing methods.

Glossy Bell Comparisons

Here, we provide the novel view synthesis and normal comparisons of our proposed RGS with 3DGS-DR.

Interactive Comparisons

Here we display interactive videos comparing our proposed RGS with 3DGS-DR. See more details in the paper.

Select an object below:


Interactive visualization. Hover or tap to move the split.