K-Buffers: A Plug-in Method for Enhancing Neural Fields with Multiple Buffers

TL;DR

This paper introduces K-Buffers, a plug-in technique that uses multiple buffers and a fusion network to significantly improve rendering performance in neural fields for 3D vision and graphics.

Contribution

The paper presents a novel plug-in method, K-Buffers, that enhances neural field rendering by leveraging multiple buffers and a fusion network, applicable to existing scene representations.

Findings

Improves rendering speed and quality for neural point fields.

Enhances 3D Gaussian Splatting with better rendering performance.

Effective across multiple neural scene representations.

Abstract

Neural fields are now the central focus of research in 3D vision and computer graphics. Existing methods mainly focus on various scene representations, such as neural points and 3D Gaussians. However, few works have studied the rendering process to enhance the neural fields. In this work, we propose a plug-in method named K-Buffers that leverages multiple buffers to improve the rendering performance. Our method first renders K buffers from scene representations and constructs K pixel-wise feature maps. Then, We introduce a K-Feature Fusion Network (KFN) to merge the K pixel-wise feature maps. Finally, we adopt a feature decoder to generate the rendering image. We also introduce an acceleration strategy to improve rendering speed and quality. We apply our method to well-known radiance field baselines, including neural point fields and 3D Gaussian Splatting (3DGS). Extensive experiments demonstrate that our method effectively enhances the rendering performance of neural point fields and 3DGS.