One-Shot Refiner: Boosting Feed-forward Novel View Synthesis via One-Step Diffusion

TL;DR

This paper introduces a novel framework combining a dual-domain perception module and a diffusion-based refinement network to improve high-fidelity novel view synthesis from sparse images, overcoming resolution and consistency limitations of previous ViT-based methods.

Contribution

It proposes a unified approach integrating high-resolution feature handling and diffusion refinement to enhance 3D view synthesis quality, a novel combination not previously explored.

Findings

Outperforms existing methods across multiple datasets.

Maintains high-frequency details during synthesis.

Ensures consistent structures across views.

Abstract

We present a novel framework for high-fidelity novel view synthesis (NVS) from sparse images, addressing key limitations in recent feed-forward 3D Gaussian Splatting (3DGS) methods built on Vision Transformer (ViT) backbones. While ViT-based pipelines offer strong geometric priors, they are often constrained by low-resolution inputs due to computational costs. Moreover, existing generative enhancement methods tend to be 3D-agnostic, resulting in inconsistent structures across views, especially in unseen regions. To overcome these challenges, we design a Dual-Domain Detail Perception Module, which enables handling high-resolution images without being limited by the ViT backbone, and endows Gaussians with additional features to store high-frequency details. We develop a feature-guided diffusion network, which can preserve high-frequency details during the restoration process. We introduce a unified training strategy that enables joint optimization of the ViT-based geometric backbone and the diffusion-based refinement module. Experiments demonstrate that our method can maintain superior generation quality across multiple datasets.