UniTriSplat: A Unified 3D Gaussian Splatting Framework with Uniform Spherical Rasterization for Universal Cameras

2026-06-29

AI Summary

Existing 3D Gaussian splatting methods suffer from inconsistent solid-angle sampling and degraded performance when handling heterogeneous camera models—such as perspective, fisheye, and omnidirectional—due to their reliance on camera-specific rasterization strategies. This work proposes UniTriSplat, the first unified 3D Gaussian splatting framework that generalizes across diverse camera types. By modeling Gaussians on the unit sphere using HEALPix, the method establishes an equal-area sampling grid aligned with angular image resolution and directly derives forward rendering and gradient propagation in the spherical domain. Additionally, a HEALPix-aware SSIM loss is introduced to preserve spherical neighborhood structure. UniTriSplat significantly enhances cross-camera generalization, maintaining geometric fidelity and high-quality rendering across scenes ranging from narrow fields of view to full 360° panoramas.

Abstract

Existing 3D Gaussian Splatting (3DGS) frameworks rely on camera-specific rasterization, suffering from inconsistent solid-angle sampling and degraded performance across heterogeneous camera models (e.g., perspective, fisheye, omnidirectional). To address this limitation, we propose UniTriSplat, a unified 3DGS framework for universal cameras that reformulates Gaussian splatting on the unit sphere via HEALPix discretization. Leveraging the equal-area property of HEALPix, we construct a spherical sampling grid aligned with the angular resolution of input images. We derive the forward rendering and gradient propagation of Gaussians directly in the spherical radian domain, yielding uniform optimization behavior from narrow-FoV images to full 360-degree panoramas. To enhance perceptual reconstruction quality, we additionally introduce a HEALPix-aware SSIM loss that respects spherical neighborhood structure. Extensive experiments across diverse camera models demonstrate that UniTriSplat consistently improves cross-camera generalization while preserving geometric fidelity and rendering quality.

Problem

• 3D Gaussian Splatting
• universal cameras
• spherical rasterization
• cross-camera generalization
• solid-angle sampling

Innovation

• 3D Gaussian Splatting
• HEALPix
• spherical rasterization
• universal cameras
• cross-camera generalization