Date: Monday, March 17 2008
Speaker: Georgios Ouzounis
Room and time: 5161.0267 (Bernoulliborg), 16.00
Title: Hyper-connectivity, operators and applications
Hyper-connectivity is a recent development in the field of mathematical morphology and the theory of connectivity. It is expressed, like ordinary connectivity, by means of classes which associate to specific operators. Hyper-connected operators however, lack cetain properties which leads to a weaker mathematical foundation as compared to ordinary connected operators. In the specific case of k-level hyper-connectivity, we bypass the practical implications of these constraints and provide the means to extract hyper-connected regions that do not associate to any partitioning scheme. This allows us to implement hyper-connected attribute filters that employ an improved scheme for connected component clustering using intensity rather than spatial criteria. The filters are implemented on the Max-Tree algorithm using the mask-based framework from our earlier work. We demonstrate their potential on 3D medical data-sets and compare our findings with the results obtained from existing connected attribute filters.
Colloquiumcoördinator is Prof.dr. M. Aiello, Dr. M. Biehl (e-mail: M.Biehl@cs.rug.nl)