LongSplat: Online Generalizable 3D Gaussian Splatting from Long Sequence Images

TL;DR

LongSplat introduces an online, real-time 3D Gaussian reconstruction framework for long-sequence images, enabling efficient incremental updates and high-quality novel view synthesis with reduced Gaussian counts.

Contribution

The paper presents LongSplat, a novel streaming update mechanism with Gaussian-Image Representation for efficient, online 3D Gaussian reconstruction from long image sequences.

Findings

Achieves state-of-the-art efficiency-quality trade-offs in real-time view synthesis.

Reduces Gaussian counts by 44% compared to existing methods.

Enables continuous, incremental updates without high memory or computational costs.

Abstract

3D Gaussian Splatting achieves high-fidelity novel view synthesis, but its application to online long-sequence scenarios is still limited. Existing methods either rely on slow per-scene optimization or fail to provide efficient incremental updates, hindering continuous performance. In this paper, we propose LongSplat, an online real-time 3D Gaussian reconstruction framework designed for long-sequence image input. The core idea is a streaming update mechanism that incrementally integrates current-view observations while selectively compressing redundant historical Gaussians. Crucial to this mechanism is our Gaussian-Image Representation (GIR), a representation that encodes 3D Gaussian parameters into a structured, image-like 2D format. GIR simultaneously enables efficient fusion of current-view and historical Gaussians and identity-aware redundancy compression. These functions enable online reconstruction and adapt the model to long sequences without overwhelming memory or computational costs. Furthermore, we leverage an existing image compression method to guide the generation of more compact and higher-quality 3D Gaussians. Extensive evaluations demonstrate that LongSplat achieves state-of-the-art efficiency-quality trade-offs in real-time novel view synthesis, delivering real-time reconstruction while reducing Gaussian counts by 44\% compared to existing per-pixel Gaussian prediction methods.