Scientific Visualization
Our group has been pursuing a number of issues in scientific
visualization, including isosurface extraction, direct volume
rendering, hierarchical data structures, visualization of curvilinear
and other irregularly gridded data, visualization of time-varying
grids, and interpolation between sample
data points. We can visualize experimental or simulated scalar
or vector data in multi-dimensional spaces.
Research is under the direction of
Professor Jane Wilhelms
and
Professor Allen Van Gelder .
Graduate students who have been involved in the project include
Jonathon Gibbs,
Kwansik Kim,
Thomas Raffill,
Jiannhwa Sun,
Paul Tarantino,
Vivek Verma.
This research is currently being funded by
NAS/NASA-Ames Research Center (NAS 2-991 and NAG 2-1239)
and the
National Science Foundation. (CCR-9503829 and CDA-9115268)
A list of our
publications in this area is available.
Examples of Current Research
Isosurfaces
Isosurfaces are 2D surfaces that can be
extracted from 3D (or higher dimensional) sample volumes using a variety
of methods. Our particular interests have included studying
topological issues in extracting surfaces, making extraction faster,
and extracting isosurfaces from a variety of grid types.
The left image is of the rectilinear data set "hipip", courtesy of
Scripps Institute ,
and the right is of the curvilinear data set
Blunt Fin.
Direct Volume Rendering
In direct volume rendering, all the data in the volume can contribute
color and opacity to the final image. Our research includes
direct volume rendering work of irregular and multi-grids,
of time-varying grids,
and of very large volumes using hierarchies. The two molecular images
on the left are again of "hipip" courtesy of
Scripps Institute .
The first image uses coherent projection (hardware Gouraud shading)
and the next compares four direct volume rendering methods.
Next, the
Space Shuttle Launch Vehicle
consists of nine separate and intersecting grids with about a million
data samples and the density field was imaged
using a new software method on a reality engine with 4 processors
in 47 seconds elapsed time.
Finally, the
Unsteady Langley Fighter is a tetrahedral dataset.
Visualization of Curvilinear and Other Irregular Grids
Curvilinear grids can be thought of as rectilinear grids that were
warped in space. They are often used when the grid must represent
curved surfaces, as in computational fluid dynamics. They present
particular problems because points vary greatly in proximity,
shapes of cells defined by corner points vary, and multiple
overlapping grids can be found. Two curvilinear data sets are
the
Blunt Fin and the
Space Shuttle Launch Vehicle, which is really nine
overlapping curvilinear grids.
Other irregular grids may consist of tetrahedral or other polyhedral
cells, such as the
Unsteady Langley Fighter .
Hierarchical Visualization
Hierarchies offer the advantages of summarizing large and complex data
sets. We first built hierarchies over grids for isosurface extraction,
using a space-efficient BON (branch-on-need) octree method.
More recently, we have developed direct volume rendering software
that builds an octree over regular data sets. Each node stores
various kinds of information, including an approximation of the underlying
data and data importance. Hierarchies can offer great speed
advantages because they restrict computation to regions actually
contributing to the image or to regions of importance. They also
recognize when the approximate model is close enough to the actual
data, and use it when appropriate. The images show data drawn
at various levels of the hierarchy.
Volume Decimation
Rendering highly complex models can be time and space prohibitive,
and decimation is an important tool in providing simplifications.
A decimated model may replace the original or provide level-of-detail
approximations. We have developed and evaluated methods for rapidly
decimating volumetric data defined on a tetrahedral grid, comparing
results using both direct volume rendering and isosurface rendering.
Geometric and data-based error metrics are used. The left image shows
Unsteady Langley Fighter
before decimation (using direct volume rendering)
and the right images show it after 95% of the vertices have been decimated.
Vector Data Visualization
This shows vector data from the
Liquid Oxygen Post (left)
and
Blunt Fin data sets (center and right).
These are stills from a program that uses color table
animation to visualize flow data. The color of the moving dots
indicates the energy, the curve indicates flow direction, and the
speed of the motion indicates velocity.
Animations
This animation (Quicktime, 1.24Mb)
shows a fly-by of the
Unsteady Langley Fighter , a tetrahedral dataset.
This animation (Quicktime, 1.52Mb)
shows a fly-by of the
Space Shuttle Launch Vehicle , a multi-grid with
nine curvilinear data sets.
This animation (Quicktime, 2.4Mb)
shows an animation of the
Impinging Jet , a time-varying, multi-grid,
curvilinear data set.
This animation (Quicktime, 2.6Mb)
shows a close-up animation of the
Impinging Jet , a time-varying, multi-grid,
curvilinear data set,
also shown in the previous animation.
.
Scientific Visualization
Isosurfaces are 2D surfaces that can be
extracted from 3D (or higher dimensional) sample volumes using a variety
of methods. Our particular interests have included studying
topological issues in extracting surfaces, making extraction faster,
and extracting isosurfaces from a variety of grid types.
The left image is of the rectilinear data set "hipip", courtesy of
Scripps Institute ,
and the right is of the curvilinear data set
Blunt Fin.
In direct volume rendering, all the data in the volume can contribute
color and opacity to the final image. Our research includes
direct volume rendering work of irregular and multi-grids,
of time-varying grids,
and of very large volumes using hierarchies. The two molecular images
on the left are again of "hipip" courtesy of
Scripps Institute .
The first image uses coherent projection (hardware Gouraud shading)
and the next compares four direct volume rendering methods.
Next, the
Space Shuttle Launch Vehicle
consists of nine separate and intersecting grids with about a million
data samples and the density field was imaged
using a new software method on a reality engine with 4 processors
in 47 seconds elapsed time.
Finally, the
Unsteady Langley Fighter is a tetrahedral dataset.
Curvilinear grids can be thought of as rectilinear grids that were
warped in space. They are often used when the grid must represent
curved surfaces, as in computational fluid dynamics. They present
particular problems because points vary greatly in proximity,
shapes of cells defined by corner points vary, and multiple
overlapping grids can be found. Two curvilinear data sets are
the
Blunt Fin and the
Space Shuttle Launch Vehicle, which is really nine
overlapping curvilinear grids.
Other irregular grids may consist of tetrahedral or other polyhedral
cells, such as the
Unsteady Langley Fighter .
Hierarchies offer the advantages of summarizing large and complex data
sets. We first built hierarchies over grids for isosurface extraction,
using a space-efficient BON (branch-on-need) octree method.
More recently, we have developed direct volume rendering software
that builds an octree over regular data sets. Each node stores
various kinds of information, including an approximation of the underlying
data and data importance. Hierarchies can offer great speed
advantages because they restrict computation to regions actually
contributing to the image or to regions of importance. They also
recognize when the approximate model is close enough to the actual
data, and use it when appropriate. The images show data drawn
at various levels of the hierarchy.
Rendering highly complex models can be time and space prohibitive,
and decimation is an important tool in providing simplifications.
A decimated model may replace the original or provide level-of-detail
approximations. We have developed and evaluated methods for rapidly
decimating volumetric data defined on a tetrahedral grid, comparing
results using both direct volume rendering and isosurface rendering.
Geometric and data-based error metrics are used. The left image shows
Unsteady Langley Fighter
before decimation (using direct volume rendering)
and the right images show it after 95% of the vertices have been decimated.
This shows vector data from the
Liquid Oxygen Post (left)
and
Blunt Fin data sets (center and right).
These are stills from a program that uses color table
animation to visualize flow data. The color of the moving dots
indicates the energy, the curve indicates flow direction, and the
speed of the motion indicates velocity.