東北大学 大学院情報科学研究科 情報基礎科学専攻 計算機構論分野
(東北大学 工学部 電気情報物理工学科 情報工学コース)
青木・伊藤(康)研究室

IEEE Student Award The Best Paper Prize

A Study on Novel View Synthesis with NeRF from Equirectangular Images

Shintaro Ito (Tohoku University) , Koichi Ito (Tohoku University) , Takafumi Aoki (Tohoku University)
電気関係学会東北支部連合大会, August 2024.
Abstract

Neural Radiance Fields (NeRF) can be used for reconstructing scenes to render images at arbitrary viewpoints from a set of images, which requires numerous perspective images. The use of equirectangular images captured by a 360-degree camera makes it possible to render large-scale scenes with fewer input images since equirectangular images can capture the entire surrounding environment in a single shot. On the other hand, converting equirectangular images to perspective images is necessary for using them as input for NeRF. This conversion may introduce the strong distortion in the images, resulting in poor rendering quality. In this paper, we propose a preprocessing method that enables the use of equirectangular images as input to NeRF. We demonstrate the effectiveness of the proposed method through experiments on a public dataset.

戻る