Skip to Main content Skip to Navigation

Systèmes de représentation multi-échelles pour l'indexation et la restauration d'archives médiévales couleur

Abstract : Our work has been motivated by a practical problem aiming at seeking objects within images and in particular in image databases of medieval frescos. The developed tools will have to help the art historians in their daily work in enabling retrieving of similar images or objects in the group of the images they have. This study is consequently within the framework of content based image retrieval. In order to solve this problem, the traditional methods are ineffective because they characterize the images as a whole without taking the arrangement of the objects which composes them into account. Suggested method describes the image by a pyramidal graph. The images are considered to be made up complex objects in relationship. A algorithm computes several coherent segmentations isolating the objects and theirs parts. Then, the adjacency graphs of the various levels are linked in order to obtain the pyramidal graph of the image. Within this high-level representation structure, we characterize each region most correctly as possible using its color, its texture and its shape. Various algorithms are finally proposed to use these descriptions of the objects, the spatial and the pyramidal informations of the representation in order to determine the similarity between images or to detect similar objects within the database. Many experimental examples illustrate this manuscript and show that this approach is overall more powerful than the existing systems.
Document type :
Complete list of metadata
Contributor : Julien Dombre Connect in order to contact the contributor
Submitted on : Wednesday, June 9, 2004 - 5:50:51 PM
Last modification on : Wednesday, October 20, 2021 - 3:22:08 AM
Long-term archiving on: : Friday, April 2, 2010 - 8:01:44 PM


  • HAL Id : tel-00006234, version 1



Julien Dombre. Systèmes de représentation multi-échelles pour l'indexation et la restauration d'archives médiévales couleur. Interface homme-machine [cs.HC]. Université de Poitiers, 2003. Français. ⟨tel-00006234⟩



Record views


Files downloads