TD-NeRF: Novel Truncated Depth Prior for Joint Camera Pose and Neural Radiance Field Optimization

TL;DR

TD-NeRF introduces a novel method for jointly optimizing camera poses and neural radiance fields from monocular images, effectively utilizing depth priors with a truncated normal sampling, coarse-to-fine training, and robust constraints to improve accuracy and convergence.

Contribution

It proposes a new approach that explicitly leverages monocular depth priors for joint camera pose and NeRF optimization, addressing noise and local minima issues.

Findings

Outperforms prior methods in pose and depth accuracy

Faster convergence in training process

Produces more accurate 3D reconstructions

Abstract

The reliance on accurate camera poses is a significant barrier to the widespread deployment of Neural Radiance Fields (NeRF) models for 3D reconstruction and SLAM tasks. The existing method introduces monocular depth priors to jointly optimize the camera poses and NeRF, which fails to fully exploit the depth priors and neglects the impact of their inherent noise. In this paper, we propose Truncated Depth NeRF (TD-NeRF), a novel approach that enables training NeRF from unknown camera poses - by jointly optimizing learnable parameters of the radiance field and camera poses. Our approach explicitly utilizes monocular depth priors through three key advancements: 1) we propose a novel depth-based ray sampling strategy based on the truncated normal distribution, which improves the convergence speed and accuracy of pose estimation; 2) to circumvent local minima and refine depth geometry, we introduce a coarse-to-fine training strategy that progressively improves the depth precision; 3) we propose a more robust inter-frame point constraint that enhances robustness against depth noise during training. The experimental results on three datasets demonstrate that TD-NeRF achieves superior performance in the joint optimization of camera pose and NeRF, surpassing prior works, and generates more accurate depth geometry. The implementation of our method has been released at https://github.com/nubot-nudt/TD-NeRF.