×

Flooding and segmentation. (English) Zbl 1073.68870

Goutsias, John (ed.) et al., Mathematical morphology and its applications to image and signal processing. Proceedings of the 5th international symposium ISMM, Palo Alto, CA, USA, June 26–29, 2000. Boston: Kluwer Academic Publishers (ISBN 0-7923-7862-8/hbk). Comput. Imaging Vis. 18, 189-198 (2000).
Summary: Flooding is at the heart of morphological segmentation. The properties of floodings, defined as upper levelings are first studied. If \(g\) is a flooding of \(f\), its catchment basins are unions of the catchment basins of \(f\). For this reason, it is possible to construct multiscale segmentations associated to families of increasing floodings of a given reference function. Various useful families of floodings are presented, together with ways to combine them. We then show how traditional segmentation techniques belong to this framework if one floods with a uniform level of water and with markers. Replacing this uniform flooding by others permits to favour the segmentation of regions having some characteristics such as large size or contrast. Finally the introduction of fuzzy markers establishes a continuum between traditional segmentation with markers and pure multiscale segmentation.
For the entire collection see [Zbl 0971.00025].

MSC:

68U10 Computing methodologies for image processing