Indexation Sémantique des Vidéos : cas des Journaux Télévisés Arabes.

Résumé

The work carried out in this thesis is part of the semantic indexing of videos and focuses more particularly on Arabic news videos. Our contribution at the theoretical level consists in proposing an indexing approach allowing to go from a low-level digital representation to a semantic description of video content by exploiting the embedded text as a source of information. This approach is composed of two modules. A low-level processing module introduces a phase of the extraction of the key frames and a step of detection and recognition of the embedded text. The second module allows the extraction of semantic information according to multi-faceted modeling: conceptual, event and theme. At the experimental level, our contribution is manifested by the development of the SISAVIN system. This includes all the modules needed for the semantic indexing of television news, ranging from the extraction of key frames to the generation of final indexes in the form of xml files. The results obtained during the evaluation are generally satisfactory for the various methods proposed and show that the SISAVIN system is capable of providing effective and relevant indexing for Arabic news videos.

Citer ce document

Accès au document

Ce document est hébergé sur son dépôt institutionnel d'origine.

Auteur(s)

Statistiques

Consultations : 2

Téléchargements : 0