Efficient ray sampling for radiance fields reconstruction

Shilei Sun*, Ming Liu, Zhongyi Fan, Qingliang Jiao, Yuxue Liu, Liquan Dong, Lingqin Kong

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

2 Citations (Scopus)

Abstract

Accelerating the training process of neural radiance field holds substantial practical value. The ray sampling strategy profoundly influences the convergence of this neural network. Therefore, more efficient ray sampling can directly augment the training efficiency of existing NeRF models. We propose a novel ray sampling approach for neural radiance field that improves training efficiency while retaining photorealistic rendering results. First, we analyze the relationship between the pixel loss distribution of sampled rays and rendering quality. This reveals redundancy in the original NeRF's uniform ray sampling. Guided by this finding, we develop a sampling method leveraging pixel regions and depth boundaries. Our main idea is to sample fewer rays in training views, yet with each ray more informative for scene fitting. Sampling probability increases in pixel areas exhibiting significant color and depth variation, greatly reducing wasteful rays from other regions without sacrificing precision. Through this method, not only can the convergence of the network be accelerated, but the spatial geometry of a scene can also be perceived more accurately. Rendering outputs are enhanced, especially for texture-complex regions. Experiments demonstrate that our method significantly outperforms state-of-the-art techniques on public benchmark datasets.

Original languageEnglish
Pages (from-to)48-59
Number of pages12
JournalComputers and Graphics (Pergamon)
Volume118
DOIs
Publication statusPublished - Feb 2024

Keywords

  • Efficient ray sampling
  • Neural radiance field
  • Training efficiency

Fingerprint

Dive into the research topics of 'Efficient ray sampling for radiance fields reconstruction'. Together they form a unique fingerprint.

Cite this