Skip to Main content Skip to Navigation

Une approche incrémentale à base de processus coopératifs et adaptatifs pour la segmentation des images en niveaux de gris

Fabrice Bellet 1
TIMC - Techniques de l'Ingénierie Médicale et de la Complexité - Informatique, Mathématiques et Applications, Grenoble - UMR 5525
Abstract : Computer vision is mainly aimed at the extraction of pertinent data, given a large amount of raw and low level information included in grey level pictures. Two kind of primitives are built, the first one is called contour or edge, and corresponds to picture areas where a significative intensity change is visible for human eyes, and the second one is called region, and corresponds to a group of pixels sharing some common grey level properties, like homogeneity for example. A careful information management is allowed thanks to segmentation task distribution into several independant entities, spatially located into the picture, being in charge of a specific edge or region primitive to be segmented, and building them in an incremental way, one pixel after the other. This approach proposes an original view point to implement cooperation between these working entities. Both kind of segmentation processes work at the same time in the same picture in parallel, taking advantage from each others properties. The edge detector creates intances of new region detectors to validate the edge under construction, and the region detector creates instances of edge detectors to topologically limit its growth. This processes set develops into a tree of cooperative entities, linked by parent-child relation. Such an approach is well suited to local adaptation, because each primitive is built by a particular instance of a generic detector. Detection parameters can then be adjusted independently from other detectors. The cooperation is effective because it is part of the decision process. The system is finaly implemented into an anonymous task scheduler. It simulates a pseudo parallelism, and uses essentially well known mechanisms usually dedicated to computer operating systems.
Complete list of metadatas

Cited literature [75 references]  Display  Hide  Download
Contributor : Thèses Imag <>
Submitted on : Thursday, February 19, 2004 - 9:59:14 AM
Last modification on : Thursday, November 19, 2020 - 3:56:08 PM
Long-term archiving on: : Friday, April 2, 2010 - 7:57:30 PM


  • HAL Id : tel-00004867, version 1



Fabrice Bellet. Une approche incrémentale à base de processus coopératifs et adaptatifs pour la segmentation des images en niveaux de gris. Interface homme-machine [cs.HC]. Institut National Polytechnique de Grenoble - INPG, 1998. Français. ⟨tel-00004867⟩



Record views


Files downloads