Segmentation of range images
Tomáš Pajdla,
Czech Technical University, Faculty of Electrical Engineering, CZ-121 35 Praha 2, Karlovo náměstí 13, Czech Republic; {pajdla,hlavac,sara}@vision.felk.cvut.cz
Václav Hlaváč,
Czech Technical University, Faculty of Electrical Engineering, CZ-121 35 Praha 2, Karlovo náměstí 13, Czech Republic; {pajdla,hlavac,sara}@vision.felk.cvut.cz
Radim Šára,
Czech Technical University, Faculty of Electrical Engineering, CZ-121 35 Praha 2, Karlovo náměstí 13, Czech Republic; {pajdla,hlavac,sara}@vision.felk.cvut.cz
Abstract
The range image (RI, gives the distance from an observer) segmentation task is motivated by 3D object reconstruction from multiple RI of an object. To simplify the task of object reconstruction a particular RI is segmented to regions that can be precisely described by a parametric model. The output of segmentation is the list of regions with attached surface models and the image describing the topology of regions. When a sequence of range images is once segmented, the correspondence between particular views can be established and 3D model of the object can be recovered. The aim of this article is to introduce a field of research activity which is applicable not only to computer vision problems.
Keywords : differential geometry, discontinuity, range image, segmentation
Pour citer cet article
Tomáš Pajdla, Václav Hlaváč & Radim Šára, «Segmentation of range images», Acta Stereologica [En ligne], Volume 13 (1994), Number 2 - Proceedings of the sixth European congress for stereology - Part three - June 1994, 459-464 URL : https://popups.uliege.be/0351-580x/index.php?id=4343.