Abstract : We present an efficient method to segment large 3d images that takes the whole 3-dimensional structure into account. We use a Union-Find data structure to record and maintain the necessary informations during the segmentation process. This structure leads to a volume labelling and facilitates interactive process by linking directly voxels to the volume they belong to. Moreover the global integration of all the data enhances the accuracy of statistical merging criteria.
https://hal.inria.fr/inria-00099034
Contributeur : Jens Gustedt
<>
Soumis le : mercredi 22 décembre 2010 - 22:56:46
Dernière modification le : jeudi 11 janvier 2018 - 06:26:07
Document(s) archivé(s) le : mercredi 23 mars 2011 - 02:22:16
