Visualization and Networking Toolkits with Wavelets Gordon Erlebacher

Visualization and Networking Toolkits with Wavelets Gordon Erlebacher Florida State University David A. Yuen University of Minnesota May 5 -10, 2002 ACES 2002, Maui, HW

Beyond Wavelets Ø E. Candes (Caltech) D. Donoho (Stanford University) Ø Ø Wavelets (point singularities) Curvelets (curve singularities) Surflets (surface singularities) Beamlets (edge detection in images) Ø Early development: § Inefficient compared to wavelet transforms § Compare to wavelets 10 years ago May 5 -10, 2002 ACES 2002, Maui, HW 2

Curvelet Transform Original Orig + noise Based on ridgelets Wavelet Transform Curvelet Transform Donoho & Huo wavelet constant Multiscale Do & Vetterli 2001 May 5 -10, 2002 ACES 2002, Maui, HW 3

Beamlets e. g. , Edge Extraction Hierarchical beam basis 256 x 256 = 65 k pixels May 5 -10, 2002 900 beamlets ACES 2002, Maui, HW 4

Fault extraction via beamlets Image from Regenauer & Yuen 2002 Shear zones on venus Feature extraction via wavelets May 5 -10, 2002 San Andreas fault ACES 2002, Maui, HW Ice ridges and grooves on Europa Microstructural image of mylonitc shear zone 5

Returning to wavelets … May 5 -10, 2002 ACES 2002, Maui, HW 6

May 5 -10, 2002 ACES 2002, Maui, HW 7

May 5 -10, 2002 ACES 2002, Maui, HW 8

Urgent Needs Ø 3 D data compression Ø Better data representation Ø Methods for feature quantification Ø Efficient automatic feature extraction Ø Next two slides illustrate this using § 2 D thermal convection at increasing Ra § 3 D thermal convection at high Ra May 5 -10, 2002 ACES 2002, Maui, HW 9

Temperature field, 2 D grid: 3400 x 500 Ra = 3× 107 Ra = 3× 108 Ra = 109 Ra = 1010 May 5 -10, 2002 ACES 2002, Maui, HW 10

May 5 -10, 2002 ACES 2002, Maui, HW 11

Wavelet-Based Toolkit Ø Visualization requires the ability to compute auxiliary variables § Given velocity, density, pressure, compute temperature transport § Compute the time-derivative of some variable Ø Variables must be computed on a time-dependent adaptive grid Ø Need to compute variables over § User-specified spatial region § User-specified scales § With a range of thresholds Ø Need to compute statistical quantities May 5 -10, 2002 ACES 2002, Maui, HW 12

Advanced Visualization Amira: www. amiravis. com Ø General-purpose visualization and 3 D reconstruction software Ø Ideally suited for 3 D datasets: scalar and vector fields Ø Advanced volume visualization Ø Object-Oriented Ø Advanced manipulators § users can interact directly with the data Ø Extensible by the user with developer version Ø Flowchart-based Ø Harnesses hardware of commodity graphics cards May 5 -10, 2002 ACES 2002, Maui, HW 13

Wavelet Thresholding Module development in Amira Wavelets: 1. 2% of coefficients Flowchart GUI Full resolution May 5 -10, 2002 ACES 2002, Maui, HW 14

Wavelet Thresholding Feature identification May 5 -10, 2002 ACES 2002, Maui, HW 15

Remote Visualization Ø Data could be computed, accumulated, stored, analyzed, and visualized at different locations Ø Data is stored in many databases around the world Ø Users collaborate § In the same location § At distributed locations Ø Need toolkits to simplify access, analysis, and visualization of the data in a transparent fashion!! May 5 -10, 2002 ACES 2002, Maui, HW 16

Video Streaming with wavelets Visualization CORBA/SOAP Server GUI Ipaq Frame Wavelet transform Ipaq Frame Color animations at 4 frames/sec on Ipaq (320 x 200) and 802. 11 b wireless network Encode May 5 -10, 2002 Visualization Wavelet transform Decode ACES 2002, Maui, HW 17

May 5 -10, 2002 ACES 2002, Maui, HW 18

SERVICES (slide provided by Fox) (A) Community Contributed Services (research). (B) Earth. Scope Provided Services. Earth. Scope does not have to produce; can access existing (distributed) products. - Visualization Service: (commercial, open source) Needs: 3 D, 4 D, overlay, georeferenced. - Registration Service: different datasets into common reference system [e. g. , GIS]. - Simple data mining tools: exist, new research mining tools will eventually become contributed as a standard service. - Data Aggregation Service: combine different datasets to form meta-sets. - Higher level Application Data Structure Service: (e. g. , interpolation of Finite Element mesh). May 5 -10, 2002 ACES 2002, Maui, HW 19

Interactive Web Querying Another Grid Service Ø Data Maps Ø 3 D data stored in various remote sites Ø Data can be queried for § Statistical information of primitive or derived variables (hook up wavelet calculator to this system) Ø User interface optimized for handheld devices May 5 -10, 2002 ACES 2002, Maui, HW 20

Map of data Histogram Two-way flow of information!! May 5 -10, 2002 ACES 2002, Maui, HW 21

Wireless Speeds Present and Near Future Ø Present: 802. 11 b § Range: 150 m § 10 Mbit/sec Ø 1 st quarter 2002: 802. 11 a § Range: 150 m § 54 Mbit/sec § Not compatible with 802. 11 b Ø 3 rd quarter 2002: 802. 11 g § Range: N/A § 54 Mbit/sec § Compatible with 802. 11 b!! May 5 -10, 2002 ACES 2002, Maui, HW 22

OQO: true mobile computing? Fall 2002 Ø Ø • • Up to 1 GHz Crusoe chip 256 Mbytes memory 10 Gbyte hard disk Touchscreen USB/Firewire Windows XP 4” screen May 5 -10, 2002 ACES 2002, Maui, HW 23

Conclusions Ø Size of datasets is exploding Ø Wavelets help to § § Compress the data (1/100) Visualize the data Analyze the data Communicate between centers Ø Wireless communication promises § Better access to field data § Ubiquitous access to data using pocket devices May 5 -10, 2002 ACES 2002, Maui, HW 24

The End May 5 -10, 2002 ACES 2002, Maui, HW 25

Beamlets Ø Ø is to look at tracks (not cracks) and fault-like strtuctures produced in laboratory experiments. There is a laboratory experiment done with glass recently to look for faults and tracks which span from the micron to 3 cm range the effective aspect-ratio is around 2 x 10**4 x 1 something you cannot do in numerical experiments so easily but beamlets would be a definite application. May 5 -10, 2002 ACES 2002, Maui, HW 26

May 5 -10, 2002 ACES 2002, Maui, HW 27

Beamlets Ø Objective: extract edges information from a noisy image Ø Edges are expressed as a series expansion in “beamlets” : Ø Issues: develop fast transforms to and from beamlet space May 5 -10, 2002 ACES 2002, Maui, HW 28

ANALYSIS FLOWS (KNOWLEDGE PATHS) Schematic of Slide Shown Earlier By Geoffry Fox (Monday afternoon, March 25). DATA SOURCE Raw data DATA Raw data STRUCTURE Flows Vary branches (Web) Service SERVICES EARTHSCOPE FRAMEWORK Data Mining, Imaging/ Analysis, Visualization MIDDLE TIER USER May 5 -10, 2002 iterations Portal ACES 2002, Maui, HW 29

DATA STRUCTURES *Earth. Scope has all Data Types: point vector time series polygon/surface matrix volume & time (4 D) * Plus Higher Level Application Data Structure e. g. , F. E. mesh, F. D. volume, Kirchhoff imaging volume ES/IT ACTION ITEM (Needs to be done fairly early): (A) Define Earth. Scope Data Structures. - Broad definitions common to all. - Foundation for an Earth. Scope Framework. (B) Define Earth. Scope Framework. - Provides commonality and communication between services. - Define up to the level of Earth. Scope observable data. - Build upon this basic definition to describe particular datasets (done by discipine). May 5 -10, 2002 ACES 2002, Maui, HW 30

Grid Services (Fox et al. 2002, Concurrency & Practice 2001) Ø Collaborative Portal § XML-based § Secure Ø Coupling of § Multi-scale numerical simulations / observational data § 4 D space-time domain (visualization) § Data mining § Efficient I/O mechanisms § Computational Steering § Databases May 5 -10, 2002 ACES 2002, Maui, HW 31

Wireless Portal May 5 -10, 2002 ACES 2002, Maui, HW 32

Web Services G. Fox Ø Suscribe/Publish Model Ø Based on current standards § XML, XSL, schemas § Developed with Java § Room for alternate web-ready languages, i. e. , Python Ø Peer to Peer structure Ø Offers wide range of services § Computation § Collaborative § Visualization May 5 -10, 2002 ACES 2002, Maui, HW 33

Grid Services (Fox et al. 2001) Ø Collaborative Portal § XML-based § Secure Ø Coupling of § § § Multi-scale numerical simulations / observational data 4 D space-time domain (visualization) Data mining Efficient I/O mechanisms Computational Steering Databases May 5 -10, 2002 ACES 2002, Maui, HW 34