Efficient 3D Gaussian splatting surface rendering for reflective objects

Multi-view image reconstruction of high-quality 3D scenes with realism rendering is a key task in the fields of virtual reality, film and television special effects, and digital twins. In recent years, 3DGS, as an alternative to NeRF, has achieved a good balance between real-time rendering and image quality by virtue of its differentiable tile rasterization and explicit representation. However, 3DGS still faces challenges in modelling reflective objects associated with high-frequency viewing angles. In this paper, we propose a Based Physical Rendering enhancement method, which combines deferred rendering strategy, binds the BRDF physical attributes and decouples the illumination and geometry on each Gaussian to achieve accurate reflection modelling. Meanwhile, an anisotropic spherical Gaussian function is introduced to replace the spherical harmonic to model complex indirect illumination and perform visibility tracking on the extracted mesh to enhance reflection realism. In addition, this paper proposes a Gaussian pruning strategy based on multi-view importance and normal-depth consistency loss to effectively eliminate redundant Gaussians, improve the geometric reconstruction accuracy, and alleviate the noise problem in normal estimation. Experimental results show that the method has good results in reflective object rendering quality, geometry reconstruction accuracy and rendering efficiency.

Authors

Institutions

Publication Details

Published
2025-09-19
DOI
https://doi.org/10.1117/12.3077433
Primary Topic
Computer Graphics and Visualization Techniques
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

Efficient 3D Gaussian splatting surface rendering for reflective objects

Xiaojun Cheng, Peiyao Xie
Computer Graphics and Visualization Techniques
article

Efficient 3D Gaussian splatting surface rendering for reflective objects

Xiaojun Cheng, Peiyao Xie
article en

Abstract

Multi-view image reconstruction of high-quality 3D scenes with realism rendering is a key task in the fields of virtual reality, film and television special effects, and digital twins. In recent years, 3DGS, as an alternative to NeRF, has achieved a good balance between real-time rendering and image quality by virtue of its differentiable tile rasterization and explicit representation. However, 3DGS still faces challenges in modelling reflective objects associated with high-frequency viewing angles. In this paper, we propose a Based Physical Rendering enhancement method, which combines deferred rendering strategy, binds the BRDF physical attributes and decouples the illumination and geometry on each Gaussian to achieve accurate reflection modelling. Meanwhile, an anisotropic spherical Gaussian function is introduced to replace the spherical harmonic to model complex indirect illumination and perform visibility tracking on the extracted mesh to enhance reflection realism. In addition, this paper proposes a Gaussian pruning strategy based on multi-view importance and normal-depth consistency loss to effectively eliminate redundant Gaussians, improve the geometric reconstruction accuracy, and alleviate the noise problem in normal estimation. Experimental results show that the method has good results in reflective object rendering quality, geometry reconstruction accuracy and rendering efficiency.

Jingdezhen Ceramic Institute (CN)
Openalex Percentile: Top 54%
Computer Graphics and Visualization Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.