Efficient and high quality clustering [Elektronische Ressource] = Effiziente und hochqualitative Clusterbildung / von Iurie Chiosa

Computergraphik undMultimediasystemeEfficient and High Quality ClusteringEffiziente und HochqualitativeClusterbildungvom Fachbereich Elektrotechnik und Informatikder Universit¨at Siegenzur Erlangung des akademischen GradesDoktor der Ingenieurwissenschaften (Dr.-Ing.)genehmigte DissertationvonIurie ChiosaSiegen, Juli 20101. Gutachter: Prof. Dr. Andreas Kolb2. Gutachter: Prof. Dr. Mario BotschTag der mu¨ndlichen Pru¨fung: 23 September 2010Gedruckt auf alterungsbestan¨ digem holz- und s¨aurefreiem Papier.Revision 1.2.4AbstractClustering,asaprocessofpartitioningdataelementswithsimilarproperties,isanessentialtask in many application areas. Due to technological advances, the amount as well as thedimensionalityofdatasetsingeneralissteadilygrowing.Thisisespeciallythecaseforlargepolygonal surface meshes since existing 3D geometry acquisition systems can nowadaysprovide models with up to several million triangles. Thus, fast and high-quality data andpolygonal mesh processing becomes more demanding.To deal with such a huge and diverse heap of clustering problems efficient algorithmsare required. At the same time the resulting clustering quality is of highest importance inalmost all situations. Thus, identifying an optimal tradeoff between efficiency and qualityis crucial in many clustering tasks.Fordataclusteringtasksingeneralaswellasmeshclusteringapplicationsinparticular,k-means like techniques or hierarchical methods are used most often.