tnm093 – practical data visualization and...

9/12/2013

1

Scientific Visualization GroupLinköping University

TNM093 – PRACTICAL DATA VISUALIZATION

AND VIRTUAL REALITY

Timo Ropinski

CONTENT

This part of the course deals with„An interdisciplinary branch of science, primarily concerned with the visualization of three dimensional phenomena, such as architectural, meteorological, medical, biological systems. The emphasis is on [realistic] rendering of volumes, surfaces, [illumination sources], and so forth, perhaps with a dynamic (time) component.“

TNM093 - Practical Data VisualizationIntroduction

1

9/12/2013

2

SCIENTIFIC VS. INFORMATION VISUALIZATION

Direct Volume Rendering

Streamlines

Line Integral Convolution

GlyphsIsosurfaces

SciVis

Scatter Plots

Parallel Coordinates

Node-link Diagrams

InfoVis[Verma et al.,Vis 2000]

[Hauser et al.,Vis 2000]

[Cabral & Leedom,SIGGRAPH 1993]

[Fua et al., Vis 1999]

[http://www.axon.com/gn_Acuity.html]

[Lamping et al., CHI 1995]

[Tamara Munzner]

TNM093 - Practical Data VisualizationBasics

2

COURSE OVERVIEW

1. Introduction and Basics2. The Data to be Visualized3. Scalar Visualization Techniques4. Vector Visualization Techniques5. Color Mapping6. Volume Rendering Theory7. Volume Rendering Praxis8. Perception and Cognition9. Volumetric Haptics10.Medical Visualization Challenges11.(Reserved)


3

9/12/2013

3

LITERATURE

Schroeder et al.: The Visualization Toolkit, An Object‐Oriented Approach To 3D Graphics, Prentice Hall.

Hansen, Johnson: The Visualization Handbook, Elsevier Academic Press.

Additional books will berecommended duringthe course


4

THE SCIVIS GROUP AT LIUStaff

Mark Eric Dieckmann

Björn Gudmundsson

Karljohan Lundin Palmerius

Claes Lundström

Timo Ropinski

Anders Ynnerman

PhDs & Research Engineers

Alexander Bock

Daniel Jönsson

Sathish Kottravel

Stefan Lindholm

Tan Khoa Nguyen

Erik Sundén


5

9/12/2013

4

SCIVIS RESEARCH AREAS

Volume Rendering

Medical Visualization

Haptics

Visualization in other Domains


6

SCIVIS WEBSITE


7

http://scivis.itn.liu.se

9/12/2013

5

WORKING WITH THE GROUP

Student co‐assistant opportunities

Master Thesis projects in interesting environments

PhD positions


8

CURRENT RESEARCH


9

9/12/2013

6

APPLICATION AREAS


10

GOOGLE EARTH


11

9/12/2013

7

VISIBLE HUMAN PROJECT

[National Library of Medicine]


12

HUMPBACK WHALE BEHAVIOR


13

[Ware et al., IEEE CG&A 2006]

9/12/2013

8

DISCIPLINES IN SCIENTIFIC VISUALIZATION

Data Domain‐based visualization

• Medical data ⇒ Volume Visualization

• Flow data ⇒ Flow Visualization

• GIS data ⇒ Geovisualization

Other fields• Microscopic data (molecular physics),

• Macroscopic data (astronomy)

• Extremely large data sets

• etc. …


14

MODERN APPLICATION AREAS

Medical visualization

Geovisualization• 3D GIS

• Oil and gas industry

Industrial design

Biological visualization

…

[Baumann 2000]


15

9/12/2013

9

MEDICAL VISUALIZATION

Image‐based diagnosis

Pre‐operative planning

Illustrative techniques forpresentation andcommunication

[Voxelman Project] [Preim 2001]

[Amide]


16

GEOVISUALIZATION – CITY VISUALIZATION

Different application areas• Tourism

• Location‐based services

• Planning & presentation

• …


17

9/12/2013

10

GEOVISUALIZATION - SEISMIC VISUALIZATION

Problem• Find optimal location for well drilling(injection and extraction)

Approach• Leverage human intuition

• Couple simulationand visualization


18

VISUALIZATION TOOLKITS


19

9/12/2013

11

VTK – THE VISUALIZATION TOOLKIT 1/2 Open‐source, freely available software system for 3D computer graphics, image processing, and visualization

VTK consists of a C++ class library and several interpreted interface layers including Tcl/Tk, Java, and Python• scalar, vector, tensor, texture, and volumetric methods

• polygon reduction, mesh smoothing, cutting, contouring, and Delaunay triangulation

• information visualization

• 3D interaction widgets[www.vtk.org]


20

VTK – THE VISUALIZATION TOOLKIT 2/2

[www.vtk.org]


21

9/12/2013

12

AMIRA

Commercial toolkit for visualization in life sciences

[www.amiravis.com]


22

MEVIS LAB

[www.mevislab.de]


23

9/12/2013

13

VOREEN

[http://www.voreen.org]


24

DATA STRUCTURES


25

9/12/2013

14

DATA SOURCES

Two main sources for data

• Real world (Measurement)collected through measurements or observations

• Theoretical world (Simulation)computed on the basis of mathematical models

Size of data depends on• Number of samples

• Number of parameters per sample

• Number of possible values per parameter

TNM093 - Practical Data VisualizationData

26

MEASUREMENT & SIMULATION

Measurement

Simulation


27

9/12/2013

15

DATA SPACE

Data space is the space in which the data is measured or simulated

The data is acquired for different observation points

• e.g., sensors in a sensornetwork


28

DATA SPACE - DIMENSIONALITY

Dimensionality (1D, 2D, 3D,…) determines the number of independent variables• Discrete vs. continuous(4D: running clock vs. epoche‐based)

Scientific visualization: Inherent spatial domain• 2D/3D data space given• Examples: medical data, flow simulation data, GIS‐data, ...

Information visualization: No inherent spatial reference• Abstract data, spatializing through visualization• Examples: data bases, stocks, …


29

9/12/2013

16

DATA SPACE - INFLUENCE

Three different areas of influence exist forobservation points• Punctual influence:just at the location of the observation point

• Local influence:in its vicinity

• Global influence:for the entire data space

Punctual and local influence can be visualized spatially

Local influence requires interpolation


30

DATA SPACE - CONNECTIVITY

Scattered data vs. grid structures


31

9/12/2013

17

DATA CHARACTERISTICS

Data type• Characteristic of a data value(precision & semantic)

Data dimensionality

• Scalar, multivariate / multimodal

Data domain

• Range of possible values


32

CHARACTERISTICS - DATA DIMENSIONALITY

Specifies the number of components (parameters/attributes)• Scalar

• Each data element has a numeric expression

• Example: topography of terrain, expressed as 2D field containing elevations

• Vector• Each data element is considered a straight directed line with a certain length (magnitude) in nD space

• Example: direction of particle flow in a channel

• Tensor• Each data element is expressed as a (transformation) matrix

• Example: higher‐order tensor fields


33

9/12/2013

18

CHARACTERISTICS - DATA DIMENSIONALITY

The number of dependent variables at each observation point

Data DimensionalityData space

1D 2D 3D

1D y=f(x)

2D Flow simulation v(x)

3D Medical scans (CT, MRI, …)i(x)

Multimodal scans (PET/CT)i1(x), i2(x)

DTI scansFlow simulation v(x)


34

CHARACTERISTICS – DATA DIMENSIONALITYData Space /Data Dimensionality

Description Examples

value series bar chart

function graph

function over 2D height map, contour lines, false color map

2D vector field hedgehog plot, LIC

3D intensities volume rendering, iso‐surface rendering

tuple set parallel coordinates, glyphs


35

9/12/2013

19

WHY GRIDS?

Natural phenomena are often continuous (infinite resolution)

Computers are obviously restricted to work with a finite amount of data

Sampling is necessary


36

CHOOSING THE RIGHT DATA STRUCTURE

Which data organization is optimal?

• Is there a neighborhood relationship?

• How is the neighborhood information stored?

• How is navigation within the data possible?

• What calculations are possible with the data?

• Are the locations of the observation points structured?


37

9/12/2013

20

2D GRIDS

Uniform rectilinear

• Cells: rectangles

• Layout: uniform

Curvilinear• Cells: different quads

• Layout: non‐uniform, structured

Unstructured• Cells: polygons (mostly triangles)

• Layout: unstructured


38

3D GRIDS

Uniform rectilinear

• Cells: regular hexahedra

• Layout: uniform

Curvilinear• Cells: different hexahedra

• Layout: non‐uniform, structured

Unstructured• Cells: different convex polyhedra

• Layout: unstructured


39

9/12/2013

21

LOCAL INTERPOLATION

Interpolation is an approximate reconstruction of the original signal

• Input: data values given for cell corners

• Output: data value for an arbitrary point inside a cell

?


40

1D - LINEAR INTERPOLATION

Compute a new data value based on two known values


41

9/12/2013

22

2D – BILINEAR INTERPOLATION


42

3D –TRILINEAR INTERPOLATION4 linear interpolations

2 bilinear interpolations

1 trilinear interpolation


43

9/12/2013

23

ALGORITHM TYPES

Algorithm type based on underlying data

• Scalar algorithms

• Vector algorithms

• Tensor algorithms

TNM093 - Practical Data VisualizationChapter 3: Scalar Visualization

44

VOLUME RENDERING

TNM093 - Practical Data VisualizationVolume Rendering

45

9/12/2013

24

OUTLINE

Data Set 3D Rendering

in real‐time on commodity graphics hardware

Classification


46

VOLUME RENDERING

Image Plane

Image order approach:

For each pixel {calculate color of the pixel

}

Data Set

Eye


47

9/12/2013

25

GPU-BASED RAYCASTING

Has several benefits over slicing• Easy to implement

• Flexible (e.g., adaptive sampling, ERT)

• Correct perspective projection

• Native 32‐bit compositing


48

BASIC RAY SETUP / TERMINATION


49

Two main approaches:• Procedural ray/box intersection[Röttger et al., 2003], [Green, 2004]

• Rasterize bounding box[Krüger and Westermann, 2003]

Some possibilities• Ray start position and exit check

• Ray start position and exit position

• Ray start position and direction vector

9/12/2013

26

PROCEDURAL RAY SETUP/TERMINATION

Everything handled in the fragment shader

Procedural ray / bounding box intersection

Ray is given by camera positionand volume entry position

Exit criterion needed

Pro: simple and self‐contained

Con: full load on the fragment shader


50

- =

"IMAGE-BASED" RAY SETUP/TERMINATION

Rasterize bounding box front faces and back faces[Krüger and Westermann, 2003]

Ray start position: front faces

Direc on vector= (back faces−front faces)

Independent of projection(orthogonal/perspective)


51

9/12/2013

27

FRAGMENT SHADER

Rasterize front facesof volume bounding box

Texcoords are volumeposition in [0,1]

Subtract camera position

Repeatedly check forexit of bounding box


52

COMPOSITING

Emission/Absorption Maximum Intensity Projection


53

9/12/2013

28

CLASSIFICATIONMapping intensities to optical properties through transfer functions


54

COLOR MAPPING

Maps a scalar to a color

Implemented using lookup table indexed through the scalar value

)

Values stored in lookup table• RGB or HSV color components

• Alpha/opacity value

TNM093 - Practical Data VisualizationScalar Visualization

55

9/12/2013

29

MAPPING TO COLOR

To label (indicate set membership)

To depict a measurement

To represent reality

To decorate (consider data/ink ratio)

representinformation


56

MAPPING TO COLOR

Color perception influenced by fruit picking

Bad choice: color ramps (rainbow pitfall)


57

9/12/2013

30

LOOKUP TABLES IN VTK

Example


58

VtkLookupTable* lut = vtkLookupTable::New();lut->SetHueRange(0.6667, 0.0);lut->SetSaturationRange(1.0, 1.0);lut->SetValueRange(1.0, 1.0)lut->SetNumberOfColors(256)lut->Build();

vtkPolyDataMapper* mapper=vtkPolyDataMapper::New();mapper->SetLookupTable(lut);mapper->SetScalarRange(min,max)

CLASSIFICATION

Classification is performed through the transfer function

Intensities are mapped to optical properties

• Which color to be assigned to which intensities?

• Which intensities are transparent?


59

9/12/2013

31

TRANSFER FUNCTIONS IN 2D


60

TRANSFER FUNCTIONS 3D


61

9/12/2013

32

VOLUME ILLUMINATION


62

WHAT IS VOLUMETRIC SHADING? Differences to surface shading

• Light/volume interaction insteadof light/surface interaction

• Translucent materials are frequent

• Volumetric shadows occur

Key problems• How does light interact with thevolume?

• How can materials be represented?

• How can scattering and shadowingbe supported?


63

9/12/2013

33

WHY VOLUMETRIC SHADING?

Langer, Bülthoff: Depth Discrimination from Shading under Diffuse Lighting [Perception 00]:

Gribble, Parker: Enhancing Interactive Particle Visualization with Advanced Shading Models[APGV06]

„… depth discrimination under diffuse lighting is superiorto that predicted by a classical sunny day model, …“


64

EMISSION ABSORPTION MODEL

Assumption:Volume consists of small particles which are

• opaque

• non‐reflecting

• light emitting

• the only light sources in the scene


65

9/12/2013

34

LOCAL VOLUMETRIC SHADING

Illumination limitations so far

• Sample points are only light sources

• No interaction with external light

Practical solution

cr(s') cri(s',Li)i1

k

')'()'()'(),'(),()( dsscscsssTIssTsI re

s

s eBebe

e

b

s'

0L

1L

2L


66

DIFFUSE LIGHTING

Diffuse lighting effects are mostprominent in volume data

Light calculation is based onLambert‘s law

Ratio can be expressed by dot product

LL N N


67

9/12/2013

35

DIFFUSE LIGHTING

no shadingdiffuse lighting

darkdiffuse lighting

bright


68

DIFFUSE LIGHTING


69

9/12/2013

36

SPECULAR LIGHTING

Specular highlights can addrealism to certain tissues

Lighting calculation is view dependent

Can be expressed as follows

LN

RV

H

diffuse

diffuse + specular


70

SPECULAR REFLECTIONS

diffuseonly

=10 =30

=120=60 =90


71

9/12/2013

37

SPECULAR LIGHTING


72

AMBIENT LIGHTING

Add constant light in shadowed regions

Drawback: contrast reduction


73

9/12/2013

38

AMBIENT LIGHTING

ambient bright+diffuse + specular

ambient medium+diffuse + specular

ambient dark +diffuse + specular


74


75

PHONG LIGHTING

Volume Rendering

9/12/2013

39

ADDING ATTENUATION


76

PHONG SHADING + ATTENUATION

no shadingPhong shadingand attenuation

Phong shading


77

9/12/2013

40

LN

RH

GRADIENT CALCULATION

Surface normal is required for diffuse and specular illumination

The gradient is a good approxi‐mation for a surface normal

[Levoy, CG&A 1988]

L N


78

GRADIENT ESTIMATION

The gradient vector is the first‐order derivative of the scalar field

We can estimate the gradient vector using finite differencing schemes• Forward/backward differences• Central differences

partial derivativein x‐direction

partial derivativein y‐direction

partial derivativein z‐direction


79

9/12/2013

41

BACK-/FORWARD DIFFERENCES

Forward differences

Backward differences


80

FORWARD DIFFERENCES


81

9/12/2013

42

CENTRAL DIFFERENCES


82

WRAP UP

TNM093 - Practical Data Visualization83

9/12/2013

43

WRAP UP

Visualization of spatial data

Scalar visualization algorithms

GPU‐based volume rendering

Classification

Volume illumination


LITERATURE

Hansen and Johnson: The Visualization Handbook, Elsevier.

Engel, Hadwiger, Kniss, Rezk‐Salama, Weiskopf:Real‐Time Volume Graphics, AK Peters.


9/12/2013

44

ACKNOWLEDGEMENTS

These slides contain figures credited to• Christof Rezk‐Salama

• Markus Hadwiger

• Klaus Engel


tnm093 – practical data visualization and...

Documents