Novel View Extrapolation with Video Diffusion Priors

TL;DR

This paper introduces ViewExtrapolator, a novel approach that uses Video Diffusion Priors to improve the realism and clarity of extrapolated views in 3D scene synthesis, outperforming existing methods.

Contribution

It presents a new method that leverages Stable Video Diffusion priors for view extrapolation without fine-tuning, enhancing realism and broadening applicability.

Findings

Outperforms existing view extrapolation methods in realism and clarity

Works with various 3D rendering types, including point clouds and monocular videos

Requires no fine-tuning, making it data- and computation-efficient

Abstract

The field of novel view synthesis has made significant strides thanks to the development of radiance field methods. However, most radiance field techniques are far better at novel view interpolation than novel view extrapolation where the synthesis novel views are far beyond the observed training views. We design ViewExtrapolator, a novel view synthesis approach that leverages the generative priors of Stable Video Diffusion (SVD) for realistic novel view extrapolation. By redesigning the SVD denoising process, ViewExtrapolator refines the artifact-prone views rendered by radiance fields, greatly enhancing the clarity and realism of the synthesized novel views. ViewExtrapolator is a generic novel view extrapolator that can work with different types of 3D rendering such as views rendered from point clouds when only a single view or monocular video is available. Additionally, ViewExtrapolator requires no fine-tuning of SVD, making it both data-efficient and computation-efficient. Extensive experiments demonstrate the superiority of ViewExtrapolator in novel view extrapolation. Project page: \url{https://kunhao-liu.github.io/ViewExtrapolator/}.