Structuration automatique de flux vidéos de télévision

Xavier Naturel 1
1 TEXMEX - Multimedia content-based indexing
IRISA - Institut de Recherche en Informatique et Systèmes Aléatoires, Inria Rennes – Bretagne Atlantique
Abstract : Automatic television structuring is a new research topic. Its development is motivated by the growth of digital television archives. This thesis proposes a framework for television structuring, which allows to automatically segment and label television streams. The work is composed of four parts: tools definition, stream segmentation, labeling, and update. Television streams are inherently repetitive. One of the main ideas of this thesis is to take advantage of the repetations to help structuring the stream. Repetitions are especially useful to detect non-programs, because of their repetitive nature. A fast method for detecting repetitions is proposed, which can deal with large amount of video, and which uses a manually labeled reference video dataset. Together with the detection of commercial separations, the repetition detection is used as an input to the segmentation process, resulting in a partition of the stream into programs ans non-programs. The program guide is then used to label each segment, using a dynamic time warping algorithm. Eventually, a method for updating the reference video dataset is proposed, which allows to be less dependent from a manually lebeled dataset. This update step also helps to minatain a constant quality over time.
Document type :
Theses
Complete list of metadatas

https://tel.archives-ouvertes.fr/tel-00524584
Contributor : Patrick Gros <>
Submitted on : Friday, October 8, 2010 - 11:19:57 AM
Last modification on : Friday, November 16, 2018 - 1:24:09 AM
Long-term archiving on : Monday, January 10, 2011 - 11:41:52 AM

Identifiers

  • HAL Id : tel-00524584, version 1

Citation

Xavier Naturel. Structuration automatique de flux vidéos de télévision. Interface homme-machine [cs.HC]. Université Rennes 1, 2007. Français. ⟨tel-00524584⟩

Share

Metrics

Record views

470

Files downloads

332