Approche interpixel en analyse d'images : une topologie et des algorithmes de segmentation
Christophe Fiorio · HAL (Le Centre pour la Communication Scientifique Directe) · 1995
Any image segmentation process aims at obtaining the features of the image entities. Most of the existing segmentation methods take into account only the materialized elements composing the image set: pixels or voxels.We suggest to deal with border elements, localized between the pixels or voxels, in order to study the links and connections between those base elements. Thus we can define a topology closely bound to image analysis processing: the star-topology. An important feature of this topology is that all classical theorems of the IR^n classical geometry can be directly translated into our frame.We associate to this topological study an inter pixel edge detector and segmentations algorithms making a perfect cooperation edge-region. Moreover we develop an extension of the region adjacency graph, the "boudary graph" which fits as well as the above results into our general framework of combinatoric methods in image analysis.