CorresNeRF: Image Correspondence Priors for Neural Radiance Fields

TL;DR

CorresNeRF introduces a method that uses image correspondence priors to enhance neural radiance fields, significantly improving performance in view synthesis and surface reconstruction, especially with limited input views.

Contribution

The paper proposes a novel plug-and-play approach that incorporates image correspondence priors into NeRF training, improving results under challenging conditions.

Findings

Outperforms previous methods in photometric metrics

Enhances geometric accuracy in surface reconstruction

Applicable across different NeRF variants

Abstract

Neural Radiance Fields (NeRFs) have achieved impressive results in novel view synthesis and surface reconstruction tasks. However, their performance suffers under challenging scenarios with sparse input views. We present CorresNeRF, a novel method that leverages image correspondence priors computed by off-the-shelf methods to supervise NeRF training. We design adaptive processes for augmentation and filtering to generate dense and high-quality correspondences. The correspondences are then used to regularize NeRF training via the correspondence pixel reprojection and depth loss terms. We evaluate our methods on novel view synthesis and surface reconstruction tasks with density-based and SDF-based NeRF models on different datasets. Our method outperforms previous methods in both photometric and geometric metrics. We show that this simple yet effective technique of using correspondence priors can be applied as a plug-and-play module across different NeRF variants. The project page is at https://yxlao.github.io/corres-nerf.