Publication

Shape segmentation and retrieval based on the skeleton cut space

Feng, C. 2017 University of Groningen. 226 p.

Research output: ScientificDoctoral Thesis

Documents

Collections of 3D shapes are rapidly growing in popularity in many application areas. To effectively use such shapes in applications such as modeling, simulations, or 3D content creation, one needs to process them. This involves, among others, cutting a shape into its natural components (an operation known as segmentation) and finding shapes similar to a given model in a large collection (an operation known as retrieval). In this thesis, we propose novel ways to segment and retrieve complex 3D shapes based on the so-caled surface skeleton of the shape. While known for many years, such skeletons have been become practically computable only recently. Based on these advances, we show how surface skeletons can be used to characterize and analyze the shape, so that operations like segmentation and retrieval can be done efficiently, robustly, and with minimal user intervention. We compare our new segmentation and retrieval methods with state-of-the-art ones on several real-world complex 3D shapes, and show how our proposal has advantages in terms of result quality. Finally, we also propose a new method to extract such surface skeletons from 3D shapes which is much simpler to implement, and has comparable performance, to comparable state-of-the-art techniques. This way, we show how we can implement an end-to-end pipeline for shape segmentation and retrieval based solely on surface skeletons.
Original languageEnglish
QualificationDoctor of Philosophy
Awarding Institution
Supervisors/Advisors
  • Telea, Alexandru, Supervisor
  • Jalba, Andrei, Co-supervisor
  • Comba, J. L. D., Assessment committee, External person
  • Veltkamp, Remco C., Assessment committee, External person
  • Petkov, Nicolai, Assessment committee
Award date22-Sep-2017
Publisher
Print ISBNs978-90-367-9905-8
Electronic ISBNs978-90-367-9904-1
StatePublished - 2017

View graph of relations

ID: 48107693