Fast Hierarchical 3D Distance Transforms
Ansprechpartner: Nicolas Cuntz
Background
This paper describes a fast approximate approach
for the GPU-based computation of 3D Euclidean
Distance Transforms (DT), i.e. distance fields
with associated vector information to the closest
object point. Our hierarchical method works on
discrete voxel grids and uses a propagation
technique, both on a single hierarchy level and
between the levels. Using our hierarchical
approach, the effort to compute the DT is
significantly reduced. It is well suited for
applications that mainly rely on exact distance
values close to the boundary.
Our technique is purely GPU-based. All
hierarchical operations are performed on the GPU.
A direct comparison with the Jump Flooding
Algorithm (JFA) shows that our approach is faster
and provides better scaling in speed and
precision, while JFA should be preferred in
applications that require a more precise DT.
Papers
| · | Nicolas Cuntz, Andreas Kolb |
| Fast Hierarchical 3D Distance Transforms on the GPU |
| In Technical Report , Computer Graphics Group, Institute for Vision and Graphics (IVG), University of Siegen, 2006
|
| [bib] [pdf]
|
| · | Nicolas Cuntz, Andreas Kolb |
| Fast Hierarchical 3D Distance Transforms on the GPU |
| In Eurographics 2007 (short paper), 2007, pages 93-96 |
| [bib] [pdf]
|
