Skip to Main content Skip to Navigation

Construction et Présentation des Vidéos Interactives

Riad Hammoud 1
1 MOVI - Modeling, localization, recognition and interpretation in computer vision
GRAVIR - IMAG - Laboratoire d'informatique GRAphique, VIsion et Robotique de Grenoble, Inria Grenoble - Rhône-Alpes, CNRS - Centre National de la Recherche Scientifique : FR71
Abstract : The arrival of the MPEG-7 standard for videos requires the creation of high level structures representing their content. The work of this thesis approaches the automatic building of a part of these structures. As a starting point, we use the tools for segmentation of moving objects. Our objectives are then to find similar objects in the video and subsequently use the similarities between camera shots to group shots into video scenes. Once these structures have been built, it is easy to provide video visualization tools for the end users which permit interactive navigation like jumping to the next shot or scene containing a person. The main difficulty lies in the great variability of observed objects: changes in point of view, scales, collusions, etc. The principal contribution of this thesis is the modeling of the variability of observations by a mixture of densities based on the Gaussian mixture theory. This modeling captures various intra-shot appearances of a tracked object and considerably reduces the number of low-level descriptors to be indexed by each tracked object. The proposed formulation led to an implementation designed for different applications: matching of tracked object models represented by Gaussian mixtures, initial building of categories of all objects present in a video by a non-supervised classification technique, extraction of characteristic views and use of detected similar objects for grouping shots into scenes. Keywords: Hyperlinked video, MPEG-7, Object recognition and classification, Variability modeling, Gaussian mixture models, Interactive video navigation, Video structure.
Document type :
Complete list of metadata
Contributor : Team Perception <>
Submitted on : Thursday, April 7, 2011 - 2:35:23 PM
Last modification on : Monday, December 28, 2020 - 3:44:02 PM
Long-term archiving on: : Friday, July 8, 2011 - 2:30:08 AM


  • HAL Id : tel-00584071, version 1




Riad Hammoud. Construction et Présentation des Vidéos Interactives. Interface homme-machine [cs.HC]. Institut National Polytechnique de Grenoble - INPG, 2001. Français. ⟨tel-00584071⟩



Record views


Files downloads