RayGauss: Volumetric Gaussian-Based Ray Casting for Photorealistic Novel View Synthesis

TL;DR

RayGauss introduces a physically consistent Gaussian-based volumetric rendering method for novel view synthesis, enabling high-quality, artifact-free renderings with efficient differentiable ray casting of irregular kernels.

Contribution

It presents a novel Gaussian-based formulation for radiance and density, and a differentiable ray casting algorithm for irregularly distributed Gaussians, improving rendering quality and scene adaptation.

Findings

Achieves superior rendering quality compared to state-of-the-art methods.

Maintains reasonable training times and achieves 25 FPS inference speed.

Effectively avoids splatting artifacts with a novel Gaussian formulation.

Abstract

Differentiable volumetric rendering-based methods made significant progress in novel view synthesis. On one hand, innovative methods have replaced the Neural Radiance Fields (NeRF) network with locally parameterized structures, enabling high-quality renderings in a reasonable time. On the other hand, approaches have used differentiable splatting instead of NeRF's ray casting to optimize radiance fields rapidly using Gaussian kernels, allowing for fine adaptation to the scene. However, differentiable ray casting of irregularly spaced kernels has been scarcely explored, while splatting, despite enabling fast rendering times, is susceptible to clearly visible artifacts. Our work closes this gap by providing a physically consistent formulation of the emitted radiance c and density {\sigma}, decomposed with Gaussian functions associated with Spherical Gaussians/Harmonics for all-frequency colorimetric representation. We also introduce a method enabling differentiable ray casting of irregularly distributed Gaussians using an algorithm that integrates radiance fields slab by slab and leverages a BVH structure. This allows our approach to finely adapt to the scene while avoiding splatting artifacts. As a result, we achieve superior rendering quality compared to the state-of-the-art while maintaining reasonable training times and achieving inference speeds of 25 FPS on the Blender dataset. Project page with videos and code: https://raygauss.github.io/