Texture-Based Flow Visualization in
Augmented and Virtual Reality Environments
Paul Benölken Holger Graf
Fraunhofer Institute for Computer Graphics
Fraunhofer Str. 5
64283 Darmstadt, Germany
paul.benoelken | holger.graf | email@example.com
Virtual and Augmented Reality allows a new way of result exploration of numerical simulations, its analysis and
interpretation by immersing the user into the data sets and/or by keeping the relationship and context to the real
environment. Up to date only a few approaches targeting the deployment of augmented reality during the
analysis stage have been published. New texture based visualization tools emerged and conciliate a better
understanding of the flow fields. This paper describes our approach for integrating texture based visualization
methods within a mixed reality set-up by introducing a ‘low cost’, fully 3D interactive post processing unit for
CFD data sets.
flow visualization, texture based algorithms, virtual reality, augmented reality.
The numerical analysis and visualization of flow
phenomena has become an indispensable part of
many different research and application fields. VR
based solutions as the pioneering work of Bryson
[Brys92] or the flow field analysis presented in
[Schu99] have shown how to ease the understanding
and interpretation of complex 3D data sets.
For visualization vector fields vector fields, glyphs,
path-, stream, streak lines or particle animations are
commonly used. Nevertheless these methods are
either limited to local vector field analysis or provide
a rather coarse spatial resolution. Hence even in
combination with advanced
techniques relevant flow characteristics may be
missed without any deeper knowledge of the flow
field. This problem is addressed by texture-based
visualization methods, which provide means for
displaying the global characteristics of a flow field in
an intuitive way with a sufficient resolution and
Virtual reality systems increase the efficiency during
the analysis stage but still lack the integration with
real environments. Here AR offers a challenging and
promising technology enabling the engineer to
environment in which he works.
In this paper we present our solution for integrating a
texture-based visualization method into a virtual and
augmented reality environment. Our application
combines traditional methods for visualizing vector
fields with the Oriented Line Integral Convolution
approach introduced by Wegenkittl et.al.[Wege96].
The presented system is based on a ‘low cost’
solution of a client server architecture and runs on a
results to the real
Permission to make digital or hard copies of all or part of
this work for personal or classroom use is granted without
fee provided that copies are not made or distributed for
profit or commercial advantage and that copies bear this
notice and the full citation on the first page. To copy
otherwise, or republish, to post on servers or to redistribute
to lists, requires prior specific permission and/or a fee.
WSCG POSTERS proceedings
WSCG’2004, February 2-6, 2004, Plzen, Czech Republic.
Copyright UNION Agency – Science Press
2. Related Work
The visualization of flow fields has been subject of
numerous developments and publications. Hauser et
al [Haus02] give an extensive overview and
categorization of existing solutions and recent
Alternatively to integral objects, texture-based or
dense integration methods as the spot noise method,
introduced by van Wijk [Wijk91] were successfully
applied for visualizing vector fields. Line integral
convolution (LIC), first published by Cabral and
Leedom [Cabr93] is a very popular and powerful
method for visualizing dense vector fields.
With their Fast LIC method Stalling and Hege
[Stal95] obtained a significant improvement in LIC
performance by exploiting the coherence along
streamlines. An additional improvement was
achieved by the Integrate&Draw method [Pere98],
which produces images with more contrast and a
time saving of ~ 50% compared to Fast-LIC.
Forsell and Cohen [Fors95] extended LIC for
curvilinear surfaces with animations techniques,
directional and magnitude information. The LIC
extension presented by Teitzel et al. [Teit97] works
on 2D unstructured grids as well as on triangulated
surfaces in 3D. With their OLIC approach
Wegenkittl et al. [Wege97] address the problem of
visualizing the orientation of the flow. Extensions
were made for speeding up this method (FROLIC)
and for animating the flow via the internet [Berg00].
For extending LIC to 3D Interrante [Inte97] presents
a combined approach of direct volume rendering and
LIC. Rezk-Salama et al. [Resk99] use 3D textures
mappings for achieving interactive performance
A significant performance gain was achieved with
the image based flow visualization technique (IBFV)
introduced by van Wijk [vanWijk 2002]. Laramee et al.
[Lara03] recently extended this method for flow field
visualizations on surfaces. However this technique
has not yet been integrated into an AR or VR
In fact, research efforts for interactive exploration of
simulation results for flow fields in an AR set-up just
started to evolve. To our knowledge, the generation
of texture based visualization methods for an
interactive exploration of flow fields in a mixed
reality set-up has so far not been presented. Most
applications are placed within the area of medicine,
architecture, edutainment or cultural heritage. An
overview can be found e.g. in [Azum01].
The closest work to this publication has been
investigated within the ARVIKA project [ARVI]. To
overcome the limitations of current simulation
packages namely the loss of the relationship to the
real environment Regenbrecht
[Rege02] make use of augmented reality to lead ‘on-
site’ inspections of simulation results within an
3. Texture based Flow Visualization
Beside the conventional technique for interactively
computing and displaying 3D stream curves (lines,
tubes, faces) isosurfaces and cross sections (see
[Beno03]) we implemented an algorithm for
generating and mapping flow textures onto polygonal
surfaces like cross sections or isosurfaces.
As the orientation of the underlying vectors is an
important feature for
understanding the 3D flow characteristics we
modified and extended the fast oriented LIC
(FROLIC) method for polygonal surfaces.
The original FROLIC algorithm operates on regular
grid structures where the topological information is
given implicitly. Hence, for efficiently processing
triangulated surfaces additional spatial data structures
are required. In our implementation we used a binary
hash table for storing the surface edges and enabling
a fast access on adjacent triangles. Furthermore the
triangles are mapped into local Euclidian space for
achieving an efficient point localization, rasterization
and integration. A 4th order Runge-Kutta scheme is
applied for performing the numerical integration. As
the computation of streamlines on triangulated
surfaces implies multiple crossings of triangle
boundaries the integrator needs to check whether the
current triangle has been left or not. From the
Euclidian representation of the triangles we
computed barycentric coordinates for fast point
Beside the flow orientation within the polygonal
mesh the orientation of the 3D vectors with respect to
the surface has to be reflected for achieving a correct
spatial visualization of the flow. Thus, vectors with a
tangential orientation should cause longer traces as
orthogonal vectors. This is accomplished by
adaptively computing the filter length Li by applying
the following exponential function:
the perception and
Here |v2D| is the length of the projected 2D vector
and |v3D| denotes the 3D vector length. Lmax is
the user defined maximal filter length.
The selection of seed points is another critical issue
in computing streamlines. Especially for the oriented
LIC method the initial positions of the droplets have
to be selected carefully for avoiding overlaps, which
cause artifacts in the output image and performance
losses. We evaluated different strategies for selecting
seed points (see Table 1). For the random selection
we implemented a modified sobol scheme, which
delivered the best performance with minimal
The animation of the original FROLIC method is
either accomplished by a cyclic variation of the
streamlet intensity or by color-table animation.
Nevertheless additional filter operations have to be
applied for suppressing pulsation effects.
Our implementation for animating FROLIC images
is based on a variation of seed points for each frame.
The seed points of frame i are shifted with a constant
factor along the streamlines. These output seed
points are used as an input for the streamline
computation in the subsequent frame i+1.
For this purpose the seed point selection is prepared
within two steps: First, a set of seed points is selected
using the random selection mechanism. Afterwards, a
subset from the computed set of seed points is
selected by evaluating the neighborhood of a seed
point such that no other streamlet is found in a
The augmented reality setup we used for exploring
the flow field of a car is shown in figure 1. The user
is wearing a head mounted display with two
integrated micro cameras. The video see-thru mode is
accomplished by placing a video texture onto the
backplane of the virtual environment. An A.R.T.
Dtrack system is used for tracking the user’s head-,
PIP-, pen- and model positions with high quality.
This system uses retro-reflective markers, which are
tracked by two self-flashing infrared cameras.
In our client server architecture [Beno03], the
rendering client was running on a 1.2GHz Windows
PC with 512MB main memory, a PNY NVIDIA
Quadro 750 Graphics board. The compute server was
running on a 2GHz PC with 2GB main memory.
The interactive generation of the texture based
visualization methods can be achieved in providing
parameters, such as resolution and filter length,
transparency, number of animation images, etc., for
the manipulation of the texture based image
generation can be interactively modified using 3D
direct manipulation widgets as illustrated in Figure 3
Some results of the work described so far are shown
in the figures 1 and 2. Figure 1 shows a texture based
flow-field visualisation using a FROLIC –texture for
the representation of the vector field. This texture is
mapped onto a cutting plane, whereas figure 2 shows
the same visualization method mapped onto an iso-
The system set-up has been tested with different data
sets in different sizes and different resolutions. To
investigate the flow field interactively, a fast retrieval
of the starting (‘seed-‘) points is inherent for
streamline computations at interactive rates. The
following table (table 1) shows the performance of
the texture based streamline computations using
different strategies for setting the seed points.
Dataset ScanLine Diagonal Block Random
371ms 361 ms 380 ms 341 ms
7971 ms 7942 ms 7992 ms 7612 ms
260 ms 271 ms 270 ms 218 ms
2624 ms 2654 ms 2533 ms 2508 ms
Table 1: Seed-point selection performance
Within this paper we presented our solution for
texture based flow visualization on triangulated
surfaces. State of the art texture based visualisation
algorithm have been modified and extended for
polygonal surfaces. The system set-up for an
Figure 2: Mapping flow textures on isosurfaces.
3D manipulation widgets are used for parameters changes.
Figure 1: Texture based flow field visualization in AR setup.
The users view is displayed in the right top corner.
interactive exploration of flow fields in a mixed Download full-text
reality scenario is promising new application areas
within the engineering domain.
Faster computations will be achieved by image space
based methods. However, as the order of operations
(texturing, projection) completely differs from
conventional approaches, the integration into a
video-mixing AR environment is still an open issue.
Alternatively, higher frame rates might be achieved
by exploiting more flexible programmable graphic
boards such as the NVidia FX series. Moreover,
within an augmented reality set-up using video-
mixing technology, the used framegrabber boards
still impose a bottleneck in respect to performance
and image synchronization between the video image
and the generated virtual objects. Our future work
concentrates on the enhancement for texture-based
visualisations at interactive rates and optimize the
augmented reality set-up exploiting more hardware
features offered by the new generation of graphic
This work is part of an EU-funded project called
ViSiCADE (IST – 2000 – 28123). We would like to
thank our project partners in the consortium for the
support in the developments of this prototype.
Furthermore we would like to address the
studierstube team (esp. Dieter Schmalstieg) being
thankful for the provision of the source code and
support during the last months.
[Azum01] Azuma, R., Baillot, Y., Behringer, R.,
Feiner, S., Julier, S., MacIntyre B.:
Recent Advances in Augmented Reality.
IEEE Computer Graphics and Applications,
[Beno03] Benölken, P., Graf, H., Stork, A.:
Exploring Flow Fields in Augmented Reality
Environments. Workshop on Augmented Virtual
Reality (AVIR). Genf, Schweiz, 2003, p. 36-38
[Berg00] Berger, S., and Gröller, E. Color-table
animation of fast oriented line integral convolution
for vector field visualization. In Proceedings
WSCG 2000, pages 4-11, 2000.
[Brys92] Bryson, S., Levit, C.: ‘The virtual
windtunnel’. In IEEE Computer Graphics and
Applications, 12(4):25-34, 1992.
[Cabr93] Cabral, B., and Leedom, L. C. Imaging
vector fields using line integral convolution. In
Proc. Siggraph´93,1993, pages 263-270.
[Fors95] Forsell, L. K. and Cohen S. D., Using Line
integral Convolution for flow visualization:
Curvilinear Grids, variable-speed animation, and
unsteady flows. IEEE Transactions on
Visualization and Computer Graphics, Vol. 1, No
2, 1995, pages 133-141.
[Haus02] Hauser, H., Laramee, R. S., Doleisch H.:
State-of-the-Art Report 2002 on Flow
Visualization, Technical Report TR-VRVis-2002-
003, VRVis Research Center in Vienna, Austria,
[Inte97] Interrante, V., and Grosch, L. Strategies for
effectively visualizing 3D flow with volume LIC.
In Proc. Visualization ´97, pages 421-424, 1997.
[Lara03] Laramee R. S., Jobard B., and Hauser H.,
Image Space Based Visualization of Unsteady
On Surfaces. In Proceedings of IEEE Visualization
(Vis 2003), pages 131-138, October 19-24, 2003,
[Pere98] Perez, C., Visualizing 2D Flows: Integrate
and Draw. In D. Bartz, editor, Ninth Eurographics
Workshop on Visualization in Scientific
Computing, pages 132--142, April 1998.
[Rege02] Regenbrecht, H., Jacobsen, S.:
Augmentation of Volumetric Data in an Airplane
Cabin, Demonstration at IEEE and ACM
International Symposium on Mixed and
Augmented Reality, Darmstadt, 2002.
[Rezk99] Rezk-Salama, C., Hastreiter, P., Teitzel, C.,
and Ertl, T. Interactive exploration of volume line
integral convolution based on 3D-texture mapping.
In IEEE Visualization ´99, pages 233-240, 1999.
[Schu99] Schulz, M., Reck, F., Bartelheimer, W.,
Ertl, T.: ‘Interactive Visualization of Fluid
Dynamics Simulations in Locally Refined
Cartesian Grids’. In Proc. Visualization '99,
pages 413-416. IEEE,1999.
[Stal95] Stalling, D., and Hege, H.-C. 1995. Fast and
resolution independent line integral convolution.
In Proceedings Siggraph 1995, pages 249-256,
Los Angeles, CA, August, 1995.
[Teit97] Teitzel C., Grosso R., and Ertl T. Line
integral convolution on triangulated surfaces. In
Proc. Computer Graphics and Visualization ´97,
number8, pages 572-581, 1997.
[Wege97] Wegenkittl, R., Gröller, E., and
Purgathofer, W. Animating flow fields: Rendering
of oriented line integral convolution. In Computer
Animation ´97 Proceedings, pages 15-21, IEEE
Computer Societey, June,1997.
[Wijk91] van Wijk J.J. Spot noise-texture sysnthesis
for data visualization. In Sederberg, T., editor,
Computer Graphics (Siggraph ´91 Proceedings),
volume 25, pages 309-318, July 1991.
[Wijk02] van Wijk, J. J. 2002. Image Based Flow
Visualization. In SIGGRAPH 2002 Conference
Proceedings, Annual Conference Series, pages