1. Parallel Computing for 3D Data Visualisation and Transmission in Grid based Applications
Daniele D’Agostino
CNR - IMATI - Sezione di Genova
DISI - Università di Genova

2. Objective
Optimization of the process of isosurface extraction for huge data sets, to subsequently allow
an efficient transmission of the result for remote visualization using a compression and simplification steps and high performance techniques.

3. Multitier technology for distributed computing
GRID
Multitier technology for distributed computing
P2P networks are built on the analogy
of providing services in a community of peers.
GRID are built on the analogy of users
tapping into ubiquitous computing
and information resources,
similar to plugging into an electrical grid.
The problem that underlies GRID is coordinated resource sharing and problem solving in dynamic, multi institutional virtual organizations.

4. DATA GRID
Often large data collections are the most important community resources
54 Gigabyte
100 Terabyte
3,5 Petabyte every year
But this combination of large dataset size, geographic distribution of users and resources, and also computationally intensive analysis results in complex performance demands that are
not satisfied by any existing data management infrastructure.

5. Visualization process
Administrator:
By using a computer to combine information from dozens of scans, CT generates a three- dimensional picture
of a cross-section of the body, allowing doctors to target tumors and avoid vital structures.
One of the better instruments to represent results
obtained from data analysis
is their visualization.
For example meteorology.

6. Computational chain Data acquisition Data of phenomenon Analysis
Volumetrical data
Analysis
Visualization
Transmission

7. Result Visualization Internet 3D Result model Visualization
Even the result, with present technologies, can be huge…
3D Result model
Visualization
Simplification
Internet
Compression
…and sometimes there is need to remote visualization

8. Objective: the real time
With HPC techniques
Query and analysis
Data
3D Result
Model
Simplification
Compression
Internet
Visualization

9. Isosurface extraction
Given a scalar field f ( x, y, z, v ) an isosurface is a surface consisting of points in the space
in which f’s value is constant.
Classical approach:
Marching Cubes (Lorensen e Cline 1987)
Based on assumption:
a contour can only pass through a cell
in a finite number of ways
Idea: A case table is constructed and
enumerates all possible topological
states of a cell, given combinations
of scalar values at the cell points
But there are disadvantages...

10. Problems Such algorithm is no much efficient...
algorithm “brute force”
computational intensive
I/O bounded
…and, sometimes, it produces topological ambiguities, like wholes,
but this problem can be solved with few additional checks.

11. Optimization
For a given isovalue, only a smaller portion of cells are intersected by an isosurface
So it is suitable to use search structure like:
Adaptive Partition Trees (binary trees or octrees)
Interval Tree
K-D Trees (balanced binary trees)
If I subdivide the data I can perform this research with parallel processes

12. Simplification
Often data models are represented with more accuracy that it is necessary, especially for
objects sufficiently far from the virtual viewpoint (Hoppe 1996).
triangles
triangles
Reduction in number of geometric primitives means reduction of waiting time.

13. Parallel Simplification
Vertex decimation method (Schroeder 1992), that is simple and efficient, can be run in parallel by deleting sets of vertex that are really independent :
Super Independent Set (Franc e Skala 2000).

14. Compression
General purpose compression techniques have worse performance in respect to specific techniques.
The Stanford Bunny is composed of triangles: an ASCII files representing the uncompressed triangle table requires 133*T bit.
The connectivity may be compressed with GZIP down to 51*T, but with Edgebreaker (Rossignack 1999) less than 2*T!

15. Conclusions
The research activity is made in collaboration between the IMATI - CNR of Genova and the GVU Center of Georgia Institute of Technology.
The parallelization of the process and the interaction with the GRID
are the aims of my PhD thesis in Computer Science.

16. References In Computer Graphics, 21(4) pp.163-169, Luglio 1987.
W. E. Lorensen, H. E. Cline: Marching cubes: A high resolution 3D surface construction algorithm.
In Computer Graphics, 21(4) pp , Luglio 1987.
G. M. Nielson, B.Hamann: The asymptotic decider: Resolving the ambiguity in the marching cubes.
In Proceedings Visualization '91 - sponsored by the IEEE Computer Society, pp , 1991
J. Rossignac: Edgebreaker: Connectivity compression for triangle meshes.
In IEEE Transactions on Visualization and Computer Graphics, Vol. 5, No. 1, pp , 1999.
R. Oldfield: Summary of existing and developing data grids. Unpublished, Marzo 2002.
W. J. Schroeder, J. A. Zarge, W. Lorensen: Decimation of triangle mesh.
In ACM Computer Graphics, 26(2) pp , Luglio 1992.
M. Franc, V. Skala: Parallel Triangular Mesh Decimation.
In SCCG 2000 Conference Proceedings, Maggio 2000.
H. Hoppe: Progressive Meshes
In SIGGRAPH Conference Proceedings, pp , 1996.