Project VisualEyes Integrated Parallel Analysis

1
Project VisualEyesIntegrated Parallel Analysis
Visualization for Adaptive Simulations

• Chandrajit L. Bajaj

Computer Sciences T I C A M University of
Texas at Austin, TX
2
Interrogation of Axial Vortices

• How is the turbulent kinetic energy produced ?
• How do helical vortices develop ?
• Are the production terms of kinetic energy
  related to the large helical vortices ?
• Do the helical vortices rotate or move axially or
  remain stationary?

3
Collaborative Behavior
Shared Results
Shared Data
Visual Steering Clients
Shared Tasks
4
Interrogation of Human Joint Dynamics and Stress

• The Knee Joint

5
Varied Domains Physics
6
Analysis Visualization Paradigms
7
Coupled Domain, Computation Visualization
Data Synthesis Computation
Workstation
Workstation
Servers
Visualization, Querying Analysis
Domain Data Acquisition
8
Needs of Integrated Framework Technology

• Multi-resolution data storage, synthesis,
  representation and transmission
• Parallel computation of domain, simulation,
  visualization
• Parallel hardware independent programming
• Legacy issues integrating with large existing
  simulation codes.

256x256 gated MRI (23 timesteps)
3 error in MRI values
Compressed 76-79 per time step
Original heart data
Reduced heart data
9
Multi-resolution Error Bounded Discretizations
Error -Bounded Meshes from Images
Progressive Bit Transmission of Meshes
10
Progressive Compression Goals/Strategy

• Flexibility with respect to different classes of
  data
• surfaces reconstructed from unorganized
  point-sets
• surfaces reconstructed from planar slices of
  points (CT or laser scans)
• Encompassing a wide class of models
• open/closed surfaces
• any genus
• manifold non-manifold
• Progressive encoding of geometry
• bitwise progressive
• Progressive encoding of topology
• inter-layer intra-layer progressive
• Both lossy and lossless compressions allowed

11
The Rule The Exception

• Typically an object surface can be locally
  decomposed into

12
The Rule The Exception
Type of encoding Triangle Strip Generalized
Triangle Strip Generalized Triangle Strip With
Exception
Length Low medium high
BPT 0 1 (average) 1.2
13
Layered Decomposition

• One possible (simple and effective) way to build
  a layered decomposition is a greedy breadth first
  traversal of the surface vertices starting from
  an initial vertex/path.

14
Layered Decomposition

• Open problem (relevant especially for off-line
  compression of large models to be permanently
  stored)
• What is the best layered decomposition?
• How to obtain it?
• From multiple tests applied on actual data it
  seems that the actual compression factor is not
  very sensitive to the initial point (that is the
  greedy approach with random initial point works
  well).

15
Geometry Compression
0th order representation (explicit vertex
coordinates) Well known to be inefficient
1st order representation (D-vertex coordinate)
Used by most of the current geometry
compression schemes
2nd order representation (D2-vector
coordinates) Tested to be more efficient since
requires on average smaller vectors
Open questions Do we need a 3rd order
representation? Do we need a variable order
representation?
16
Geometry Compression
0th order representation (explicit vertex
coordinates)
1st order representation (D-vertex coordinate)
2nd order representation (D2-vector coordinates)
2nd order
17
Geometry Compression
VS
Lossy
bit-progressive
Vector coordinates interleaved at the bit level
to obtain a progressive representation of the
geometry
Vector Quantization optimized at the compression
stage for a given number of bits
Geometry takes typically 80 of the entire model
storage
18
Lossless Topology/Lossy Geometry
19
Analysis Visualization Paradigms
20
Seed Sets

• With this, we define a seed set
• A subset S of the nodes of G is a seed set of G
  if all the nodes of G are connected to S.

21
Signature Computation

• Consider a terrain of which you want to compute
  the length of each isocontour and the area
  contained inside each isocontour.

22
Signature Computation

• The length of each contour is a c spline
  function.

0

• The area inside/outside each isocontour is a
  spline function.

23
Signature Computation

• In general the size of each isocontour of a
  scalar field of dimension d is a spline function
  of d-2 continuity.
• The size of the region inside/outside is given by
  a spline function of d-1 continuity

24
User Interface - MRI of a human torso -

• The isocontour that bounds the region of interest
  is obtained by selecting the maximum of the
  gradient signature.
• In real time the exact value of each signature
  is displayed.

25
Rule-based Contouring (foot of the Visible
Human)

• The contour spectrum allows the development of an
  adaptive ability to separate interesting
  isovalues from the others.

26
Rule-based Contouring (CT scan of an engine)

• The contour spectrum allows the development of an
  adaptive ability to separate interesting
  isovalues from the others.

27
Thin Client Collaborative Visualization
Maryland
SDSC TSRI
Visual Steering Clients
UT Austin
28
Thin Client Architecture
LWU
HWQ
LWU
LWU
HWQ
LWU
HWQ
HWQ
LWULightweight Update
HWQ Heavyweight Query
29
Static Architecture
Thin Client
Services Architecture
Application Layer
Operation
Cost
Front Object
Front Gui
Lightweight
Rotation
DualPort Server
Lightweight
Translation
Front Service
Error Service
Heavyweight
Iso-Contour
Error Service
Error Service
Thin Services
Name Service
Heavyweight
StreamLines
Lightweight
Re-Color
Data Objects
Network Objects
30
Runtime Environment Communication Architecture
Kernel
Kernel
Browser
Front
SM
Front
Front
Front
Front
Front
SM
SM
Front
Kernel
Web Server
Browser
SM Session Manager
31
Integrated Parallel Framework
32
Distributed Architecture
Kernel
Dynamical Simulator
Visualization
Kernel
Contact Analysis
Part Modeler
SM
CM
SM Session Manager
CM Constraint Manager
Stress Analysis
Kernel
33
Constraint Management
Kernel
Dynamical Simulator
Visualization
Kernel
Contact Analysis
Part Modeler
SM
CM
SM Session Manager
CM Constraint Manager
Stress Analysis
Kernel
Constraints
Constraint Control
34
Graphical User Interface for Static Data

• The horizontal axis spans the scalar values ??
• Plot of a set of signatures (length, area,
  gradient ...) as functions of the scalar value ?.

• Vertical axis spans normalized ranges of each
  signature.
• White vertical bars mark current selected
  isovalues.

35
Graphical User Interface for time varying data
high
(?,t ) --gt c The color c is mapped to the
magnitude of a signature function of time t and
isovalue ?
c
magnitude
t
?

• The horizontal axis spans the scalar value
  dimension ?.
• The vertical axis spans the time dimension t .

low
36
Topological information.

• number of components per isocontour
• which isocontours merge together or split while
  modifying the isovalue.

an isocontour with three connected components two
of which are about to merge
37
Seed Set Construction