Site Navigation

Bachelor thesis

This page is about my bachelor thesis. To follow the traditional approach of making incomprehensibly long names for theses, I'm proud to announce the title of my bachelor thesis as Volume data generation from triangle meshes using signed distance functions.

Medical volume data
from http://volvis.org

Volumetric data is used in medical applications and in various other contexts. Since triangle meshes are the most common representation for geometric objects in computer graphics, the need to automatically generate volumentric data (or RAW data in some contexts) from meshes arises. A lot of techniques and methods has been developed in order to create such volumentric data, and the presented material is a survey of these techniques. Further, an application has been developed to automatically generate volumetric data from triangle meshes.

Dragon volume rendered at
two different ISO surfaces

Volumetric data from triangle meshes is a bit different from medical data by the means of the semantics of the values in the dataset. Medical data (created by a CT- or MRI-scanner) contains a density value at each VOXEL (a VOXEL is a volume element, just like a PIXEL for pictures), whereas a dataset from a triangle mesh contains the distance from each VOXEL to the closest point on the mesh. These distance values are mapped to density values later on to make the dataset viewable in most applications. But the main task in generating volumentric data from triangle meshes is to compute a signed distance field for the mesh. To do so, the mesh needs to partition space in a proper way (the surface must be closed), and this is a requirement on the mesh.

The advantage in having (discretisized) distance information per VOXEL does not only allow visualization of the objects surface, the object can also be grown and shrunk according to a user-defined ISO value. The resulting surface is called ISO surface.

Downloads

The following downloads are organized as follows: First, there is information material on the topic. Second, I will offer some model files in order to feed the converter. But of course you can use any other OFF model files that meets the requirements. Third, there are volume datasets created with the converter, you can download and view them with any volume viewer. Follow the links for some further downloads. Forth, the converter is available for download. Please regard the appropriate licenses for all files. If you want to use it, make sure you use it in a proper way.

Information material

Presentation slides. Presented on 2007-06-16.
Download PDF, 1.1 MB
Bachelor thesis. Filed at 2007-08-16.
Download PDF, 2.0 MB

Model files

You can also get some of the original versions of the models from the stanford scanning repository. To view the meshes, you can use any tool that understands the OFF model format. Try MeshLab for example.

Full view Original High quality Lower quality
The famous stanford bunny.
[856 KB] Original version, 35'286 vertices, 70'568 faces
[273 KB] HQ remeshed version, 13'199 vertices, 26'394 faces
[72 KB] LQ remeshed version, 3'405 vertices, 6'806 faces

Full view Original High quality Lower quality
The impressing stanford dragon.
[7.8 MB] Original version, 435'545 vertices, 871'306 faces
[1.5 MB] HQ remeshed version, 73'879 vertices, 147'572 faces
[698 KB] LQ remeshed version, 33'999 vertices, 67'980 faces

Full view Original High quality Lower quality
The eery stanford armadillon.
[6 MB] Original version, 165'954 vertices, 331'904 faces
[1.4 MB] HQ remeshed version, 70'817 vertices, 141'628 faces
[381 KB] LQ remeshed version, 18'387 vertices, 36'770 faces

Ico Low poly Medium poly High poly
A test sphere aquired by subdivision of an icosahedron.
[176 B] Icosahedron, 12 vertices, 20 faces
[9.4 KB] 3 subdivisions, 690 vertices, 1'280 faces
[40 KB] 4 subdivisions, 2'658 vertices, 5'120 faces
[173 KB] 5 subdivisions, 10'506 vertices, 20'480 faces

Volume data sets

Volume data sets are NOT YET AVAILABLE. You may want to take a look at VolVis to download some MRT datasets.

Tool for volume data generation

The newest version of the source code can be obtained using SVN. The code is GPLed.

svn co https://lonestar.exhale.de/svn/modelconv

Also snapshots of the CVS are ready for download.

[6.0 MB] gtk+-2.10.13-setup.exe, GTK runtime for Windows
[8.7 MB] Modelconverter -- Binary for Windows XP, includes GTKmm
[1.5 MB] Modelconverter -- Binary for Linux, requires libraries

The Windows version has been build and tested under Windows XP. It includes all necessary runtime DLLs but GTK. If you already have GTK installed on your machine (for example if you run The Gimp under Windows), you can try without reinstalling the runtime. Otherwise you need to install the GTK runtime.

The Linux-Version requires the following libraries to be installed: