内容紹介

This lecture discusses three-dimensional (3D) reconstruction in scattering media. 3D reconstruction from two-dimensional images is important in computer vision. However, images captured in scattering media, such as fog or murky water, degrade due to light scattering and attenuation caused by suspended particles. Unfortunately, conventional 3D reconstruction methods are affected by this image degradation in scattering media. This lecture presents image formation models for such degradation and proposes methods to enable 3D reconstruction in scattering media. Typical disparity- and shading-based 3D reconstruction methods, i.e., multi-view stereo and photometric stereo, are extended for scattering media with appropriate physics-based scattering models. The effectiveness of the proposed methods is evaluated on real data captured in foggy scenes and underwater.

詳細情報

刊年

2021

形態

電子化映像資料(33分10秒）

シリーズ名

情報科学領域・コロキアム ; 2021年度

注記

講演者所属: 情報科学領域

講演日: 2021年6月8日 3限

講演場所: エーアイ大講義室　(L1)

標題言語

英語 (eng)

本文言語

英語 (eng)

著者情報

Fujimura, Yuki