Metropolis-Hastings Sampling for 3D Gaussian Reconstruction

TL;DR

This paper introduces an adaptive, probabilistic sampling framework for 3D Gaussian Splatting that reduces heuristic reliance, improves efficiency, and maintains high-quality view synthesis across multiple datasets.

Contribution

It reformulates densification and pruning as a Bayesian sampling process, enabling dynamic Gaussian management without predefined scene complexity.

Findings

Reduces the number of Gaussians needed for scene representation.

Achieves faster convergence compared to traditional methods.

Maintains or slightly improves view-synthesis quality.

Abstract

We propose an adaptive sampling framework for 3D Gaussian Splatting (3DGS) that leverages comprehensive multi-view photometric error signals within a unified Metropolis-Hastings approach. Vanilla 3DGS heavily relies on heuristic-based density-control mechanisms (e.g., cloning, splitting, and pruning), which can lead to redundant computations or premature removal of beneficial Gaussians. Our framework overcomes these limitations by reformulating densification and pruning as a probabilistic sampling process, dynamically inserting and relocating Gaussians based on aggregated multi-view errors and opacity scores. Guided by Bayesian acceptance tests derived from these error-based importance scores, our method substantially reduces reliance on heuristics, offers greater flexibility, and adaptively infers Gaussian distributions without requiring predefined scene complexity. Experiments on benchmark datasets, including Mip-NeRF360, Tanks and Temples and Deep Blending, show that our approach reduces the number of Gaussians needed, achieving faster convergence while matching or modestly surpassing the view-synthesis quality of state-of-the-art models.