S92 Logo

Industrial Geometry

About   General   Events   Projects   Members   Publications   Internal


S09202 Coupling evolving level sets with curves and surfaces

This subproject focuses on active contour methods for curves and surfaces. In particular, we studied the interaction between different possible representations of geometric objects. The main results have been obtained in a close cooperation between the groups in Linz and Innsbruck. The activities can roughly be organized in four groups.

First, we studied evolution processes of T-spline level sets and applications to geometry reconstruction and image segmentation. Second, we explored the framework of coupled evolution and used it for dealing with geometric constraints and image segmentation. Third, we studied the relation between Gauß-Newton-type methods for orthogonal distance regression and evolution processes. Fourth, in a cooperation with subproject 3, we worked on the modeling, analysis and implementation of variational techniques on M-reps shape manifolds.

In addition to these core activities, we cooperated intensively with the Computational Geometry group (subproject 5) on questions related the use of higher order geometric primitives, in particular addressing the computation of the medial axis of planar domains which are bounded by general free-form curves. We also cooperated with subproject 1 on rational surfaces with special offsetting-related properties.

Results have been obtain in the following areas:
  • Evolution of T-spline Level sets, 3D Geometry Reconstruction
  • Coupled evolution
  • Gauß-Newton-type methods for orthogonal distance regression
  • Image Analysis
A detailed report of the first funding period is available here. See also the list of Publications.

(a) IMAGE:rocker IMAGE:rocker (b) IMAGE:reconstruction
Fig. 1: (a) The data and reconstructed mesh with sharp features. (b) 3D Geometry reconstructions.

IMAGE:square IMAGE:square IMAGE:square
Fig. 2: Approximation with area constraint. The area equal 0.4, 0.5 and 0.6.

IMAGE:medial axis
Fig. 3: Medial axis computation using spiral biarc approximation (Cooperation with subproject 5). The computation time was 13.25 seconds, where 13830 arcs have been used to describe the shape.




TU Graz Logo TU Wien Logo Uni Linz Logo Logo Innsbruck