Leveraging NeRF-Rendered Images for 3D Gaussian Splatting
2026-06-08 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors combine two popular techniques for creating new views of scenes: NeRF, which makes very high-quality images but is slower, and 3D Gaussian splatting (3DGS), which is faster but usually lower quality. They use NeRF to generate training images that help 3DGS work better, focusing on street scenes. This approach removes moving objects and adds extra top-down views, improving image quality while keeping fast rendering. They also use an image enhancement method to make the extra views clearer. Tests show their method improves street scene visuals without losing speed or quality.
Neural Radiance Fields (NeRF)3D Gaussian Splatting (3DGS)Novel View SynthesisStreet Scene RenderingTransient Object RemovalBird's-eye ViewDiffusion-based Image EnhancementRendering SpeedRendering Quality
Authors
Mizuki Morikawa, Yuta Shimizu, Chunyu Li, Yusuke Monno, Masatoshi Okutomi
Abstract
Neural radiance field (NeRF) and 3D Gaussian splatting (3DGS) are two mainstream approaches for novel view synthesis. They often show complementary performance, i.e., 3DGS demonstrating faster rendering speed and NeRF demonstrating higher rendering quality. Motivated by this, we propose leveraging NeRF-rendered images for 3DGS. Specifically, we target street scenes and utilize a pre-trained street-specific NeRF method to produce training images for a target 3DGS method. In our 3DGS training, NeRF-rendered images are used to remove transient objects in street-level input views and to generate bird's-eye views as additional views, inheriting the higher-quality rendering of NeRF into 3DGS. We further incorporate a diffusion-based image enhancement to improve the image quality of the additional views. Experimental results on one synthetic and two real datasets demonstrate that our proposed method improves street-scene rendering while preserving the speed of 3DGS and the quality of NeRF.