Hybrid geometry representations with applications in medical imaging and model repair

A key issue in computer aided design is the accurate and efficient mathematical representation of static and dynamic geometric models. Unfortunately there is no single design that fits all needs equally well. While e.g. parametric representations allow for fast enumeration of points on the model and thus for fast rendering, volumetric representations better support inside/outside queries and Boolean operations like intersection and union. The efficiency of geometric algorithms thus directly relates to the efficiency of the underlying data structures, a fact that becomes in particular apparent in dynamic models that need to be updated frequently. In this thesis we design and evaluate so-called hybrid geometry representations that combine the advantages of the traditional parametric, implicit and volumetric frameworks. Our goal is to selectively enhance applications by functionality that would otherwise be difficult to implement in a single representation alone. We demonstrate the applicability of hybrid models and show how applications from as diverse fields as medical imaging and CAD model repair can take advantage of this concept. In particular we turn our attention to hybrid representations that allow the user to explicitly control the topology of the geometric model. In the first part of the thesis, we examine active contour models (curves as well as surfaces) which frequently are employed in medical imaging for segmentation, pattern matching and object recognition. Here, the principal challenge is to incorporate a priori knowledge about the topology of the object of interest into the dynamic contour. We present hybrid extensions to traditional parametric (snake) and geometric (level-set) active contour models, that allow the user to explicitly control splitting and merging of the evolving contour and to efficiently incorporate topological constraints. In the second part of the thesis we investigate how hybrid geometry representations and algorithms can be successfully applied in model repair. CAD data like architectural or automotive models often contains artefacts like gaps, intersections, overlaps, and inconsistent normal orientations. Unfortunately, such "triangle soups" cannot directly be used in downstream applications which often are very particular about the topological and geometrical quality of their input. We present algorithms that combine volumetric and explicit methods to resolve these artefacts and produce high-quality, manifold and watertight reconstructions.

Grantor dc:publisher

Shaker

Year dc:date

2007

Author dc:creator

• Bischoff, Stephan Michael

Contributors dc:contributor

• Kobbelt, Leif

• info:eu-repo/classification/ddc/004
• Snake <Bildverarbeitung>
• Bildverarbeitung
• Volumenmodell
• Informatik
• Geometrieverarbeitung
• Topologiekontrolle
• Modellreparatur
• Geometry Processing
• Topology Control
• Model Repair

Statement dc:rights

• info:eu-repo/semantics/openAccess

Language dc:language

eng

Identifier

https://publications.rwth-aachen.de/search?p=id:%22RWTH-CONV-124120%22

OAI identifier oai:identifier

oai:publications.rwth-aachen.de:62556