Skip to Main content Skip to Navigation

Extraction et analyse d'objets-clés pour la structuration d'images et de vidéos

Abstract : The compact description of image and video content is currently a difficult task. We are interested in the objects that make up this content because of the representative power of these objects. After a review of the state of the art, this thesis presents a local segmentation method based on the irregular graph pyramid algorithm, which allows us to extract, using low-level features, regions of interest comparable to semantic objects. This method is used to precisely excise objects from still images, first in an interactive environment and then in an entirely automatic one. A motion estimation allows us to extend the process to videos by extracting the foreground entities from every frame. A filtering and a clustering of these entities allow us to retain only the most representative of each real object in the shot. These representations are called key-objects and key-views. The quality of the experimental results allows us to propose some future applications of our methods.
Document type :
Complete list of metadatas

Cited literature [132 references]  Display  Hide  Download
Contributor : Jérémy Huart <>
Submitted on : Tuesday, January 22, 2008 - 1:53:23 PM
Last modification on : Thursday, November 19, 2020 - 12:59:38 PM
Long-term archiving on: : Thursday, April 15, 2010 - 2:00:24 AM


  • HAL Id : tel-00212062, version 1



Jérémy Huart. Extraction et analyse d'objets-clés pour la structuration d'images et de vidéos. Interface homme-machine [cs.HC]. Institut National Polytechnique de Grenoble - INPG, 2007. Français. ⟨tel-00212062⟩



Record views


Files downloads