RoSplat: Robust Feed-Forward Pixel-wise Gaussian Splatting for Varying Input Views and High-Resolution Rendering

TL;DR

RoSplat introduces alpha normalization and a 3D regularizer to enhance pixel-wise Gaussian splatting, ensuring consistent brightness and reducing artifacts in high-resolution, multi-view synthesis.

Contribution

The paper proposes novel normalization and regularization techniques to improve Gaussian scale estimation and brightness consistency in 3D Gaussian splatting.

Findings

Significant reduction in over-brightness artifacts across varying input views.

Improved high-resolution rendering quality with fewer hole artifacts.

Enhanced baseline performance on benchmark datasets.

Abstract

Generalizable 3D Gaussian Splatting has recently emerged as an efficient approach for novel-view synthesis, enabling feed-forward synthesis from only a few input views. However, existing pixel-wise feed-forward methods suffer from over-bright renderings when the number of input views varies during inference, as well as insufficient supervision for accurate Gaussian scale estimation, which leads to hole artifacts, particularly in high-resolution renderings. To address these issues, we identify that the over-brightness is caused by the varying number of overlapping Gaussians and propose a simple alpha normalization strategy to maintain brightness consistency across different number of input views. In addition, we introduce an auxiliary 3D sampling-based regularizer to improve Gaussian scale estimation, thereby mitigating hole artifacts in high-resolution rendering. Experiments on benchmark datasets demonstrate that our method significantly improves baseline models under varying input-view and high-resolution rendering settings.