|
|
figure 2: w/o
interpolation (here: inverse cubic)
|
figure 3: w/o normal
calculation
|
[back to top]
Our Approach and Results
Ray tracing is already known for its accurate and physically correct
rendering of pictures. But is always considered as too slow to render
at interactive rates. Since our OpenRT group successfully showed that
it is possible, it was only a small step to apply it to volume
rendering.
As it can be seen from the pictures below, it was the right choice to
extend the OpenRT framework. Given a mid-sized volume modell (say
256^3) and using a resolution of 512 x 512 we are able to achieve up to
7 frames per second on a standard PC.
|
|
figure 4: Stanford bunny
(volumetric!) in environment map
|
figure 5: MRbrain mixed
with polygonal mirrors and box
|
|
|
figure 6: engine dataset
(transparent)
|
figure 7: bonsai tree with
several light sources and shadows
|
The images above are showing a small set of the possibilities with our
approach.
Figure 4 shows the famous stanford bunny, but
this
time it is actually a CT scan. To give this bunny an interesting look
we
mapped the uffizi hall on it.
Figure 5 basically demonstrates
the
combination of polyginal and volumetric datasets. Note that both types
are
actually interacting with each other through reflection and
intersection.
But refraction is also possible, as shown in
figure 6. This
allows to take a look at the inner parts of the engine, while one would
normally only see the surface. To have an idea of the surface we
blended both. The last example
(figure 7) shows another bonsai
tree with several light sources and shadows. But we also mapped an
environment map on it ... guess which one.
[back to top]
Future Work
In the future we would like to extend our framework to handle large and
time varying data sets. Extensions to allow maximum intensity
projection and
direct volume rendering seems also interesting. At the moment we are
trying
to render the data from the VHP (Visible Human Project).
[back to top]
Gallery
Please visit our new gallery with lots of pictures including most
recent
extensions:
Let me
see
...
Papers and Videos
- Demonstration video of many ray tracing effects on four different
models
[mpeg4, ~ 33MB]
- Fast and Accurate Ray-Voxel Intersection Techniques for
Iso-Surface Ray Tracing
Gerd Marmitt, Andreas Kleer, Ingo Wald, Heiko Friedrich,
and Philipp
Slusallek
in Girod, Magnor, Seidel (Eds.): Vision, Modeling, and
Visualization (VMV) 2004 Proceedings,
Stanford (CA), USA, November 16-18, 2004. pages=429-435
[bib]
[pdf]
[poster
(pdf)]
[back to top]
Students
This project belongs nowadays among the largest subgroups within the
Computer Graphics Lab. The following students are currently
involved in this project:
- Bernhard Tuerk (FoPra)
- Roman Brauchle (Research Assistant)
Former students:
- Jens Michael Weber (FoPra): Implementing a direct volume renderer
using a grid traverser
- Andreas Kleer (FoPra): Design and implementation of efficient
ray-cell intersection techniques
