Image Geometry - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Book Sections Year : 2023

Image Geometry

Abstract

This chapter reviews basic geometrical concepts relevant for omnidirectional vision. These comprise the image formation process, with a special emphasis on catadioptric cameras, and a brief discussion on how camera models approximate the image formation process. The distinction between central (or single-viewpoint) and non-central cameras is explained, together with a discussion of the respective advantages. The chapter also reviews the basic building blocks of structure-from-motion, ranging from projection and back-projection to pose and ego-motion estimation. It is then shown that line images are particularly import- ant when studying omnidirectional cameras, for calibration as well as for image match- ing through the epipolar geometry. Dense matching is usually sped up by pre-processing images during a rectification process, which is explained with a particular emphasis on omnidirectional images. The concepts discussed in this chapter are illustrated in short videos accessible on the Internet (creative commons licence CC-BY-NC-SA). URL’s are provided in the text.
Fichier principal
Vignette du fichier
image_geometry.pdf (4.17 Mo) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-03563140 , version 1 (09-02-2022)
hal-03563140 , version 2 (11-02-2022)

Licence

Identifiers

  • HAL Id : hal-03563140 , version 2

Cite

Peter Sturm. Image Geometry. Pascal Vasseur; Fabio Morbidi. Omnidirectional Vision: From Theory to Applications, ISTE, 2023, 9781789451436. ⟨hal-03563140v2⟩
112 View
380 Download

Share

Gmail Mastodon Facebook X LinkedIn More