Mise en correspondance stéréoscopique par approches variationnelles convexes ; application à la détection d'obstacles routiers

Abstract : This thesis deals with the problem of stereo matching and its application to obstacle detection. The main goal of stereo matching is to recover the depth information of a scene from a pair of left and right images taken from two different locations. It involves finding corresponding pixels in both images, leading to the so-called disparity map. Firstly, the stereo matching problem is solved by minimizing a convex objective function over the intersection of multiple constraint sets. These constraints arise from the prior knowledge and the observations. In order to obtain a smooth disparity field, while preserving edges, we consider appropriate wavelet and total variation based regularization constraints. The resulting optimization problem is solved with a block iterative method which offers great flexibility in the incorporation of several constraints. Then, to deal with illumination variations often encountered in practice, we develop a spatially varying multiplicative model that accounts for brightness changes between both images in the stereo pair. Finally, we detect obstacles from the computed depth map by performing an object segmentation based on an orientation surface criterion.
Complete list of metadatas

Cited literature [146 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00738363
Contributor : Wided Miled <>
Submitted on : Thursday, October 4, 2012 - 11:16:59 AM
Last modification on : Wednesday, April 11, 2018 - 12:12:03 PM
Long-term archiving on : Friday, December 16, 2016 - 7:37:54 PM

Identifiers

  • HAL Id : tel-00738363, version 1

Citation

Wided Souid-Miled. Mise en correspondance stéréoscopique par approches variationnelles convexes ; application à la détection d'obstacles routiers. Traitement du signal et de l'image [eess.SP]. Université Paris-Est, 2007. Français. ⟨tel-00738363⟩

Share

Metrics

Record views

393

Files downloads

675