SplatSuRe: Selective Super-Resolution for Multi-view Consistent 3D Gaussian Splatting

TL;DR

SplatSuRe introduces a selective super-resolution method for 3D Gaussian Splatting that improves multi-view consistency and detail by applying super-resolution only in undersampled, high-frequency regions based on camera pose and scene geometry.

Contribution

The paper presents a novel approach that selectively applies super-resolution in undersampled regions, enhancing 3D rendering quality without uniform SR application.

Findings

Outperforms baselines in fidelity and perceptual quality on multiple datasets.

Most significant improvements occur in localized foreground regions.

Method reduces multi-view inconsistencies and blurriness in high-resolution renders.

Abstract

3D Gaussian Splatting (3DGS) enables high-quality novel view synthesis, motivating interest in generating higher-resolution renders than those available during training. A natural strategy is to apply super-resolution (SR) to low-resolution (LR) input views, but independently enhancing each image introduces multi-view inconsistencies, leading to blurry renders. Prior methods attempt to mitigate these inconsistencies through learned neural components, temporally consistent video priors, or joint optimization on LR and SR views, but all uniformly apply SR across every image. In contrast, our key insight is that close-up LR views may contain high-frequency information for regions also captured in more distant views and that we can use the camera pose relative to scene geometry to inform where to add SR content. Building on this insight, we propose SplatSuRe, a method that selectively applies SR content only in undersampled regions lacking high-frequency supervision, yielding sharper and more consistent results. Across Tanks & Temples, Deep Blending, and Mip-NeRF 360, our approach surpasses baselines in both fidelity and perceptual quality. Notably, our gains are most significant in localized foreground regions where higher detail is desired.