Skip to Main content Skip to Navigation
Theses

Outils de traitement d'images adaptés au traitement d'images omnidirectionnelles

Abstract : In this thesis, we develop new image processing tools suited to omnidirectional vision. These new tools are based on the equivalent sphere, as defined by Geyer and Daniilidis. This is motivated by the following facts : classical image processing tools, that are suited for planar images, provide us with errors, since these methods do not take into account the peculiar geometry of omnidirectional images. Our approach give efficient methods, with the same resolution quality in the periphery or in the center of the image. This thesis is organized as follows : in a first section, we introduce what is omnidirectional image associated with a catadioptric sensor, and we recall what is the equivalent sphere. In a second section, we introduce the mathematical tools which will be used in the sequel : Fourier analysis on the sphere, spherical harmonics, spherical convolution. We then construct edge detection operators and smoothing operators based on this approach. We finally benchmark our methods on some omnidirectional images, and we discuss the advantages of our low-level processing tools against classical planar operators. These tests provide us with visual evidences of the advantages of our method, specially for a processing that is uniform on the image.
Document type :
Theses
Complete list of metadatas

Cited literature [47 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00339785
Contributor : Stéphanie Bigot-Marchand <>
Submitted on : Tuesday, November 18, 2008 - 9:27:36 PM
Last modification on : Thursday, January 11, 2018 - 6:12:26 AM
Long-term archiving on: : Monday, June 7, 2010 - 9:44:08 PM

Identifiers

  • HAL Id : tel-00339785, version 1

Collections

Citation

Stéphanie Bigot-Marchand. Outils de traitement d'images adaptés au traitement d'images omnidirectionnelles. Mathématiques [math]. Université de Picardie Jules Verne, 2008. Français. ⟨tel-00339785⟩

Share

Metrics

Record views

770

Files downloads

1233