Depth estimation from stereo cameras through a curved transparent medium

Seongwook Yoon, Taehyeon Choi, Sanghoon Sull

Research output: Contribution to journalArticle

Abstract

In this paper, we propose a novel method for estimating depth values by stereo cameras through a curved transparent medium that causes refraction. Our method takes both the surface shape of the medium and the refraction into account. We model that the rays from the stereo cameras are refracted by a curved transparent medium whose inner surface is represented by a parametric model, assuming that the medium has constant thickness. The parameters of the model are estimated using a constrained optimization simply by attaching several markers on the inner surface. The depth value is then estimated by the triangulation considering the refraction based on the model. The experimental results show that our method yields consistently high error reduction rates with respect to the baseline method without considering the refraction caused by the medium. In addition, our method provides satisfactory estimates for various shapes of the medium.

Original languageEnglish
Pages (from-to)101-107
Number of pages7
JournalPattern Recognition Letters
Volume129
DOIs
Publication statusPublished - 2020 Jan

Keywords

  • Camera calibration
  • Depth from stereo
  • Parametric surface model
  • Refraction

ASJC Scopus subject areas

  • Software
  • Signal Processing
  • Computer Vision and Pattern Recognition
  • Artificial Intelligence

Fingerprint Dive into the research topics of 'Depth estimation from stereo cameras through a curved transparent medium'. Together they form a unique fingerprint.

  • Cite this