|
Remote visualization means
interactive viewing of three dimensional scientific data sets over the
web. Because scientific data sets are in the gigabyte size range, it is
difficult to send the entire data set over the network.
Extraction, processing, network latency and rendering add up and make
the proposition of near real-time interactive visualization a challenge.
Moreover, the client will have a limited amount of memory and CPU power
for viewing and interacting with the data. Under the direction of
Dr. Kanad Ghose,
we are working to address these challenges. We are converting an
existing application for visualizing CAT scan medical data to run over Internet2. The application uses a
multithreaded version of the Marching Cubes algorithm, and the
data is organized by isovalue into chessboarded span-space buckets (see
references below).
|
|
A Fast Multithreaded Out-Of-Core Visualization Technique, by Peter Sulatycke and Kanad Ghose, in Proceedings of International Parallel Processing Symposium, 1999 (IPPS '99), pp. 569-575.
Fast Remote Isosurface Visualization With
Chessboarding, by Alisa Neeman, Peter Sulatycke, and Kanad Ghose,
in Proceedings of Eurographics Symposium on Parallel Graphics and
Visualization 2004 (EGPGV '04)
Powerpoint
slides
PDF slides
Goals for remote visualization of regular data sets:
Split application: thin
client, server does isosurface extraction, polygon formation,
calculation of normals
Measure for optimal packet
send size to reduce total number of sends over socket
Measure time for server
work, latency, client work, total
Split application: server
does isosurface extraction only, data compressed with chessboarding
Measure time for server
work, latency, client work, total
The above described work was completed in late June 2003 and submitted
to the SPIE Conference On
Visualization and Data Analysis
A second project, visualizing irregular data sets, is in the early design phase. Irregular data sets come from finite element analysis and computational fluid dynamics. These are techniques used for modeling and simulation in mechanical, aerospace and nuclear industries. For more details see the original presentation. Below is a prototype of the user interface showing a wireframe triangulation and convex hull for a 200 point data set.
|
|
|
 |
