Title: Thinking outside the
1Thinking outside the Visualization Box
- Ken Joy
- Visualization and Graphics Research Group
- Institute for Data Analysis and Visualization
- Computer Science Department
- University of California, Davis
2People
- Thanks to
- Ben Gregorski, Chris Co, Serban Porumbescu, Karim
Mahrous, Janine Bennett, Lok Hwa, IDAV, UC Davis - Mark Duchaineau, Peter Lindstrom, Valerio
Pascucci, LLNL - Josh Senecal, LLNL, Davis
- Hans Hagen, Kaiserslautern
- Bernd Hamann, UC Davis
- Note that the name of our organization has
changed.
3The Problem
This is a portion of an isosurface taken from the
final time step of a Richtmeyer-Meshkov
instability shock-tube simulation.
The simulation was generated on a 2000x2000x2000
grid. An isosurface of 460 million triangles was
generated by marching cubes.
4The Problem
5Simplification, Wavelets, Subdivision
- We simplified the mesh.
- We generated a fine mesh from a simplified base
mesh using Catmull-Clark subdivision. - Invented new Catmull-Clark wavelets to preserve
features. - Lots of publications,
6Original
716 of the coefficients
85 of the coefficients
91.6 of the coefficients
10Base Mesh
11Patches
12What did we learn?
- We worked on this data set for almost two years,
and still didnt have a interactive renderable
version. - Occlusion problems were killing us!
- Features were disappearing!
- and this was only one isovalue!!!
- Data size of this one isosurface exceeded the
size of the original data set. - We had to think outside of Multiresolution
Analysis to find a solution
13What did we do?
- Dynamic generation of the isosurface.
- Dynamic extraction based upon a 3D
longest-edge-bisection tetrahedral mesh which is
refined and coarsened depending on viewpoint,
error, and frame rate. - Algorithm depends on an innovative storage scheme
for the multiresolution data, and an occlusion
scheme, allowing real-time display of the
isosurface.
14Dynamic Generation of Isosurfaces
15Results!
- Dynamic Isosurface Generation Gregorski, et al.
IEEE Visualization 2002 - Time-Varying Isosurfaces Gregorski, et al.
TVCG, 2004 - Compression and Occlusion Gregorski, et al.
IEEE Visualization 2004, submitted.
16Lessons?
- We learned that we must think out of the box to
solve major problems. - Key words
- Large, massive, terascale
- Time varying
- Interactive, real-time
- Time critical
- Mathematically sound
17What happens if the texture is too big?
18Dynamic sensor data!
- Potentially billions of sensors
- Sensors are scattered and may move in time.
19Problems!
- Data is scattered.
- Data may move in time.
- Applications are time-varying and time-critical.
- We cannot afford to generate a mesh for each time
step. - Can we develop visualization methods that do not
depend on a fixed mesh?
20Meshless Isosurfaces
Isosplatting, Co, et al., 2003
21Meshless Isosurfaces
Isosplatting, Co, et al., 2003
22Meshless Isosurfaces
- Solves the multi-block isosurface problem
- i.e., no cracks.
Co, et al., VisSym 2004
23Lessons?
- Many mesh problems can be solved by meshless
techniques. - Instead of one algorithm for each mesh type
- Thinking outside the mesh
24Thinking Outside the Visualization Box
- Perhaps innovation in visualization can best be
achieved by Thinking outside the box. - In our research, we frequently think inside,
working on the little problems that improve the
existing algorithms. - What about new algorithms, new techniques,
new approaches?
25Think outside the Multiresolution box
- Think outside subdivision, wavelets,
splines, mesh simplification, etc. - "The key to terascale visualization is deciding
what to visualize." - The key is to focus the data exploration, not to
show absolutely everything. - Queries on Scientific Data Sets.
- Find those regions of interest to the user.
- Carr, Banff 2004 Volumetric Queries
26Think outside the Mesh
- Perhaps think Meshless
- Solves many problems!
27Think outside the Big Three
- Are there other fundamental visualization
algorithms besides Slicing, Isosurfaces, and
Volume Rendering for scalar fields
28Segmentation Techniques
Nielson and Franke, IEEE Visualization 1997
Bonnell, et al., IEEE Visualization 2000, TVCG
2003.
Mahrous, et al. TVCG 2004.
29Think outside the Heuristic
- Edelsbrunner Corollary (Banff, 2004)
- Work in the sunshine!
- Work in the world of theorems. Theorems require
deeper thinking about the subject and can show
ways to get out of the box. - Corollary Think outside linear
30Think outside the Scalar Field
- Vector Fields
- Tensor Fields
- DT-MRI Can we work with the original data?
- Multi-valued Fields
- Distributions at each data point
- Air Quality Problems
- Radiation Transport Simulations
- Distributions at each data point
- Vector distribution at each data point
- Tensor distribution at each data point
31Think outside 2D
- Three-dimensional time-varying multivalued data
exploration is HARD. Lets focus our activities
there. - What are separatrices in 3-dimensional vector
fields? - What are the topological properties of a
3-dimensional tensor field?
32Think outside Static
- The problems of the world are not static
problems! - Blood flow cannot be stopped during an MRI (on a
human, at least) - Time-varying Data
- Time Critical Data
- User-controlled visualizations.
33Think outside SciVis and InfoVis
- Isnt SciVis just InfoVis with spatial location?
- Focus Context in SciVis?
34Think outside the Pretty Picture
- AARGH!!!
- It looks good this way
- Pleasing to my eye
- Color scheme implies nothing!
- I believe that our field should be called Data
Exploration - Does it convey the correct information?
- Is it right? (Mike Kirby)
35Think outside the GPU
- Programming the GPU is Research in Programming
36Think outside the Academic Problem
- Hitching our research to someone elses driving
application and solving those problems on the
owners terms, leads us to richer visualization
research. (Fred Brooks, 1996) - Visualization is 40 papers per year. (Banks,
Banff 2004) Therefore, since 1990, the
visualization field consists of approximately 550
papers. - Where are our clients? (Lorensen, 2003)
- Is visualization a relevant field?
37Kens Eleven
- Think outside the Multiresolution box
- Think outside the Mesh
- Think outside the Big Three
- Think outside the Heuristic
- Think outside the Scalar Field
- Think outside 2D
- Think outside Static
- Think outside SciVis and InfoVis
- Think outside the Pretty Picture
- Think outside the GPU
- Think outside the Academic Problems
38HELP ME!
- Where do we need to think outside the box?
- What are the right questions to ask?
39Thank You
joy_at_cs.ucdavis.edu http//graphics.cs.ucdavis.edu
/joy