Skip to Main content Skip to Navigation

Restitution vidéo stéréoscopique maîtrisée: application à la Réalité Virtuelle

Abstract : Acquiring 3D scenes can be achieved by using a pair of video cameras (stereo rig). The acquisition process of these stereoscopic video images and their projection on 3D display systems is a wide-extending field which borrows from the technologies of virtual reality, computer vision and 3D filming. Within this wide field, we focused our research work on the projection in 3D of images captured by a fixed stereo rig, on virtual reality systems. The use of this technology was limited by numerous constraints (pertaining essentially to capture configurations and display conditions). In this PhD thesis, we detail the improvements we brought to some of the steps of the stereo transmission process, in order to control more precisely the display of stereoscopic videos and to alleviate some of the limitations mentioned above. We based our work on detailed mathematical modeling of the cameras, and of the capture and the display configurations. First, it was necessary to work on stereoscopic images as little distorted as possible. To obtain such images, we developed an algorithm that corrects distortions on these images. To ensure real-time rectification, we implemented this algorithm on GPU (Graphics Processing Unit), through the use of a technique based on reference tables. The interocular distance is a fundamental step to make a full use of virtual reality system. In spite of the fact that significant disparities exist from one user to another, an average value is often used for this parameter. In order to set correctly this critical parameter, we developed a method to calibrate the interocular distance of each user, so that the display of stereoscopic images on these systems will provide a more accurate perception. Finally, while viewers in 3D theaters sit in a defined area in front of the projection screen, allowing users to move freely is one of the main characteristics of virtual reality systems. To overcome the difficulties that occur when projecting images captured by a fixed stereo rig to a user in motion, we propose a method to improve the restitution of the depth perceived by the user, using a depth segmentation of the captured scene.
Document type :
Domain :
Complete list of metadata
Contributor : Ecole Arts et Métiers ParisTech Connect in order to contact the contributor
Submitted on : Thursday, January 28, 2010 - 8:00:00 AM
Last modification on : Friday, August 5, 2022 - 2:54:00 PM
Long-term archiving on: : Friday, September 10, 2010 - 2:56:01 PM


  • HAL Id : pastel-00005717, version 1


Fabien Goslin. Restitution vidéo stéréoscopique maîtrisée: application à la Réalité Virtuelle. domain_other. Arts et Métiers ParisTech, 2010. Français. ⟨NNT : 2010ENAM0001⟩. ⟨pastel-00005717⟩



Record views


Files downloads