A%20Framework%20for%20the%20Perceptual%20Optimization%20of%20Multivalued%20Multilayered%202D%20Scientific%20Visualization%20Methods

1
A Framework for the Perceptual Optimizationof
Multivalued Multilayered2D Scientific
Visualization Methods
PhD Thesis Daniel Acevedo Feliz B.S. Civil
Engineering, University of A Coruña, 1997 M.Sc.
Computer Science, Brown University, 2001
Thesis Committee David H. Laidlaw (CS) John F.
Hughes (CS) Leslie Welch (Psychology)
2
Research Question
What makes a multivalued scientific visualization
method effective?
3
Mars Satellite Data
H2O lightness
Fe size
Cl saturation
Kvs.Th spacing
4
Thesis Hypothesis
Measuring the perceptual capabilities of several
icon-based scientific visualization methods for
simple single-valued scalar datasets in 2D, and
combining that with subjective evaluations of
complex multilayered methods representing
multivalued datasets, we can generate a
predictive model of the perceptual properties of
a space of visualization methods.
5
Thesis Hypothesis
Measuring the perceptual capabilities of several
icon-based scientific visualization methods for
simple single-valued scalar datasets in 2D, and
combining that with subjective evaluations of
complex multilayered methods representing
multivalued datasets, we can generate a
predictive model of the perceptual properties of
a space of visualization methods.
size
spacing
lightness
saturation
6
Thesis Hypothesis
Measuring the perceptual capabilities of several
icon-based scientific visualization methods for
simple single-valued scalar datasets in 2D, and
combining that with subjective evaluations of
complex multilayered methods representing
multivalued datasets, we can generate a
predictive model of the perceptual properties of
a space of visualization methods.


7
Thesis Hypothesis
Measuring the perceptual capabilities of several
icon-based scientific visualization methods for
simple single-valued scalar datasets in 2D, and
combining that with subjective evaluations of
complex multilayered methods representing
multivalued datasets, we can generate a
predictive model of the perceptual properties of
a space of visualization methods.
DesignGoals
PerceptualProperties
VisualDimensions
Effectiveness
8
Contributions to Scientific Visualization

• Evaluation of a set of measurable perceptual
  properties to describe utility for exploratory
  scientific visualization2 IEEE Visualization
  2005 best poster award ? 3 IEEE VIS/TVCG 2006
  ?4 IEEE Visualization 2007 poster
• Quantification and definition of novel predictive
  models of the perceptual properties for our
  visual dimensions3 IEEE VIS/TVCG 2006 ? 4
  IEEE Visualization 2007 poster
• Validation of the use of visual design experts as
  evaluators of scientific visualization
  methods1 SIGGRAPH 2003 sketch ? 5 IEEE
  TVCG 2007 in review
• Evaluation of a set of methodologies to capture
  targeted critiques of visualization methods from
  visual design experts2 IEEE Visualization 2005
  best poster award ? 5 IEEE TVCG 2007 in
  review
• Evaluation of a novel framework for building
  these types of predictive models

9
Contributions to other fields

• Perceptual Psychology
• Extended previous experimental results to
  discreteicon-based cases with more complex
  stimuli
• New experimental design, quantification and
  modeling of some design factors
• Visual Design and Art
• A novel attempt to quantify the critique process
  used in art and visual design

10
Roadmap

1. Introduction
2. Related work
3. Perceptual properties
4. Perceptual studies
5. Evaluations using visual designers
6. Discussion and conclusions

11
Roadmap

1. Introduction
2. Related work
3. Perceptual properties
4. Perceptual studies
5. Evaluations using visual designers
6. Discussion and conclusions

12
Related Work

• Data Visualization
• Bertin, Cleveland, et al., Mackinlay, Card,
  Healey,
• MacEachren,
• Perceptual Psychology
• Callaghan, Landy, Bergen, Carswell and Wickens,
  Healey, Ware, Interrante,
• Visual Design and Art
• Wallschlaeger et al., Tufte,

Healey et al., 2004
Bokinsky, 2003
Watanabe et al., 1996
Landy et al., 1991
SIGGRAPH courseattendee
Van Gogh
13
Related Work Data Visualization
Frederic P. Brooks, Jr. (1996),The Computer
Scientists as a Toolsmith IICommunications of
the ACM, 39(3) 61-68

• Our success comes from our users success

Christopher Johnson (2004),Top Scientific
Visualization Research ProblemsIEEE Computer
Graphics Applications, 24(4) 13-17

• Quantify Effectiveness
• Identify Perceptual Issues

14
Roadmap

1. Introduction
2. Related work
3. Perceptual properties
4. Perceptual studies
5. Evaluations using visual designers
6. Discussion and conclusions

2 IEEE Visualization 2005 best poster award
3 IEEE VIS/TVCG 20064 IEEE Visualization
2007 poster
15
Design Factors for Exploratory Visualization

• Exploratory Visualization
• Prompt visual thinking
• Qualitative Not to search, compare, or explain
• Help scientists think about their problem

• Design Factors
• Characterize the exploratory process
• Represent data attributes
• Represent design goals

16
Design Factors for Exploratory Visualization

• Spatial Feature Resolution (?0)
• The size of the features
• Data Resolution (?1)
• The levels of data
• Saliency (?2) The composition dominance
• Perceptual Interference (?3) The cognitive
  effort to read
• Legibility (?4) SFR and DR loss when combined

17
Roadmap

1. Introduction
2. Related work
3. Perceptual properties
4. Perceptual studies
5. Spatial Feature Resolution
6. Data Resolution
7. Saliency
8. Perceptual Interference
9. Evaluations using visual designers
10. Discussion and conclusions

3 IEEE VIS/TVCG 20064 IEEE Visualization
2007 poster
18
Perceptual Studies
Visual Dimensions
lightness
saturation
orientation
size
spacing
19
Perceptual Studies
Independent Variables
size
spacing
20
Perceptual Studies
Independent Variables
Number of layers, order, and color
21
Roadmap

1. Introduction
2. Related work
3. Perceptual properties
4. Perceptual studies
5. Spatial Feature Resolution
6. Data Resolution
7. Saliency
8. Perceptual Interference
9. Evaluations using visual designers
10. Discussion and conclusions

3 IEEE VIS/TVCG 2006
22
Perceptual Studies
Experimental Setup 1
Saturation x Size x Spacing Lightness x Size x
Spacing Size x Spacing Spacing x Size
Spatial Feature Resolution Data Resolution

• Computer-based. 900x900
• Gamma-corrected. Controlled room illumination
• 6 Subjects. Within-subjects design
• No time limit, with breaks

23
Perceptual Studies
Spatial Feature Resolution
What is the size of the smallest spatialfeature
a visual dimension can represent?
min(p)max(p)
min(s)max(s)
?0(v)

2
2
24
Perceptual Studies
Data Resolution
How many different levels of data cana visual
dimension represent?
(jnds)
25
Perceptual Studies
Data Resolution
How many different levels of data cana visual
dimension represent?
Webers Law
?I/Ik
?1(v)a Ln(v) b
af(?,e1/(?1),s,s2,e1/(s1),p,p2,e1/(p1))
bg(?,e1/(?1),s,s2,e1/(s1),p,p2,e1/(p1))
frequency
spacing
size
26
Perceptual Studies
Data Resolution
an0?n1e1/(?1)n2sn3s2n4e1/(s1)n5pn6p2n7e1
/(p1)
bm0?m1e1/(?1)m2sm3s2m4e1/(s1)m5pm6p2m7e1
/(p1)
?
e1/(?1)
s
e1/(s1)
p
e1/(p1)
s2
p2
27
Perceptual Studies
Data Resolution
28
(No Transcript)
29
Roadmap

1. Introduction
2. Related work
3. Perceptual properties
4. Perceptual studies
5. Spatial Feature Resolution
6. Data Resolution
7. Saliency
8. Perceptual Interference
9. Evaluations using visual designers
10. Discussion and conclusions

4 IEEE Visualization 2007 poster
30
Perceptual Studies
Experimental Setup 2Perceptual Dominance
SAT, LIGHT, ORIENT, SIZE, SPA
Saliency Perceptual Interference
Size (3,6) x Spacing (3,6) x layers (2) x layer
order (2) x direction (2) x color (2)

• Computer-based. 900x900
• Gamma-corrected. Controlled room illumination
• 6 Subjects. Blocked within-subjects
  design Full factorial Fractional factorial
  (orthogonal array)
• 10s time limit, with breaks

31
Quantification of Dominance
Stimuli
What visual dimension do you perceive first?
t(v1), t(v2)
32
Quantification of Dominance
Overall Times
33
Quantification of Dominance
Orientation Example
34
Quantification of Dominance
Normalized Time Differences
35
Quantification of Dominance
Normalized Times



36
Quantification of Dominance
Normalized Times
37
Quantification of Dominance
Predictive Models
t(v1)-t(v2)
?2(v1,v2)
-1,1
Saliency
10
t(v1v2)-min(t(v1))
?3(v1v2)
0,1
max(t(v1))-min(t(v1))
Perceptual Interference
t(v2v1)-min(t(v2))
0,1
?3(v2v1)
max(t(v2))-min(t(v2))
38
(No Transcript)
39
Thesis Hypothesis
Measuring the perceptual capabilities of several
icon-based scientific visualization methods for
simple single-valued scalar datasets in 2D, and
combining that with subjective evaluations of
complex multilayered methods representing
multivalued datasets, we can generate a
predictive model of the perceptual properties of
a space of visualization methods
40
Roadmap

1. Introduction
2. Related work
3. Perceptual properties
4. Perceptual studies
5. Evaluations using visual designers
6. Validation
7. Single-valued visualizations
8. Multi-valued visualizations
9. Discussion and conclusions

1 SIGGRAPH 2003 sketch 2 IEEE Visualization
2005 best poster award 5 IEEE TVCG 2007 in
review
41
Evaluations using Visual Designers
42
Thesis Hypothesis
Measuring the perceptual capabilities of several
icon-based scientific visualization methods for
simple single-valued scalar datasets in 2D, and
combining that with subjective evaluations of
complex multilayered methods representing
multivalued datasets, we can generate a
predictive model of the perceptual properties of
a space of visualization methods
43
Design Factors for Exploratory Visualization

• Spatial Feature Resolution (?0)
• The size of the features
• Data Resolution (?1)
• The levels of data
• Saliency (?2) The composition dominance
• Perceptual Interference (?3) The cognitive
  effort to read
• Legibility (?4) SFR and DR loss when combined

44
Quantification of Legibility
Experimental Setup
Legibility Saliency Perceptual Interference
Two-valued datasetsSAT, LIGHT, SIZE, SPA
Size x Spacing x layers x layer order x color
(2)

• Computer-based. 900x900. Interactive
• Gamma-corrected. Controlled room illumination
• 4 Visual Design Experts. Within-subjects design
• Protocol analysis

45
Quantification of Legibility
Stimuli
46
Quantification of Legibility
Protocol

• Introduction and training
• Choose 2 values from the three-valued dataset
• Task Most salient?
• Task Legibility in the combination
• Task Legibility change from single-valued
  visualizations
• Task Maintaining legibility, modify the
  combination so - Value 1 dominates
• - Value 2 dominates
• - Both values dominate
• Task Switch values and reevaluate 6
• Task Switch methods and go to 2
• Task Design your preferred
  combination

47
Quantification of Legibility
Saliency Results
48
Quantification of Legibility
Conclusions

• Positive evaluation of saliency and interference
  models
• Could not disprove the null hypothesis
• Wrong hypothesis Experiment power
• Design freedom required to quantify design
  process
• Dataset dependence of multivalued visualizations

49
Roadmap

1. Introduction
2. Related work
3. Perceptual properties
4. Perceptual studies
5. Evaluations using visual designers
6. Discussion and conclusions

50
Framework Discussion

• Current approach
• Combine perception visual design
• Bottom-up
• Limited space
• Data independent
• Alternative approach
• Combine perception visual design
• Top-down
• Open design space Complex protocol
• Include data correlations

51
Framework Discussion

• Current approach
• Combine perception visual design
• Bottom-up
• Limited space
• Data independent
• Alternative approach
• Combine perception visual design
• Top-down
• Open design space Complex protocol
• Include data correlations

52
Thesis Hypothesis
Measuring the perceptual capabilities of several
icon-based scientific visualization methods for
simple single-valued scalar datasets in 2D, and
combining that with subjective evaluations of
complex multilayered methods representing
multivalued datasets, we can generate a
predictive model of the perceptual properties of
a space of visualization methods
53
Contributions

• Evaluation of a set of measurable perceptual
  properties to describe utility for exploratory
  scientific visualization2 IEEE Visualization
  2005 best poster award ? 3 IEEE VIS/TVCG 2006
  ?4 IEEE Visualization 2007 poster
• Quantification and definition of novel predictive
  models of the perceptual properties for our
  visual dimensions3 IEEE VIS/TVCG 2006 ? 4
  IEEE Visualization 2007 poster
• Validation of the use of visual design experts as
  evaluators of scientific visualization
  methods1 SIGGRAPH 2003 sketch ? 5 IEEE
  TVCG 2007 in review
• Evaluation of a set of methodologies to capture
  targeted critiques of visualization methods from
  visual design experts2 IEEE Visualization 2005
  best poster award ? 5 IEEE TVCG 2007 in
  review
• Evaluation of a novel framework for building
  these types of predictive models

54
New Lines of Research

• Other effectiveness factors
• Double mapping
• Second-order effects
• Accept lossy methods?
• Display size
• Interactivity

55
New Lines of Research

• Extension to 3D
• Stereo, parallax, immersion,
• Tougher cue interferences real depth!
• Transfer conclusions from methodology

56
Conclusion

• Visualization
• Discovery of facts and laws about the
  effectivetransformation of information into
  visual unitsto facilitate insight
• In this dissertation
• Part of that effectiveness comes from an
  efficient use of perceptual capabilities of
  the visual dimensions that form our
  visualization methods
• Measured the perceptual capabilities of some
  visual dimensions
• Evaluated a framework to build predictive models
  of those capabilities
• Learning the capabilities of our toolsgives us
  power over them

57
Acknowledgements

• Thesis Committee David Laidlaw, John Hughes,
  Leslie Welch
• Colleagues and friends
• Dan Keefe Cullen Jackson Jason Sobel
  Eileen Vote Andy Forsberg Tomer Moscovitch
• Joe LaViola Jian Chen
• Graduate students, faculty, and staff from -
  Browns Graphics Group - Browns
  Visualization Research Lab
• - Brown CS
• All the very patient experiments participants
• Ines, Maria, and all my family and friends
• Dedicated to my father, Luis Acevedo Martin

58
A Framework for the Perceptual Optimizationof
Multivalued Multilayered2D Scientific
Visualization Methods
PhD Thesis Daniel Acevedo Feliz B.S. Civil
Engineering, University of A Coruña, 1997 M.Sc.
Computer Science, Brown University, 2001
daf_at_cs.brown.edu
59
Publications

• Using Visual Design Experts in Critique-based
  Evaluation of 2D Vector Visualization Methods.
  Daniel Acevedo, Cullen Jackson, Fritz Drury, and
  David Laidlaw. Submitted to IEEE
  Transactions on Visualization and Computer
  Graphics.
• Modeling Perceptual Dominance among Visual Cues
  in Multilayered Icon-based Scientific
  Visualizations. Daniel Acevedo, Jian Chen, and
  David Laidlaw. IEEE Visualization 2007,
  Poster Compendium, Sacramento, CA, October 2007.
• Subjective Quantification of Perceptual
  Interactions among Some 2D Scientific
  Visualization Methods. Daniel Acevedo and David
  Laidlaw. IEEE Transactions on Visualization
  and Computer Graphics (Proceedings Visualization
  / Information Visualization), 12(5),
  September-October 2006.
• Using Visual Design Expertise to Characterize
  the Effectiveness of 2D Scientific Visualization
  Methods. Daniel Acevedo, Cullen Jackson, David
  Laidlaw, and Fritz Drury. IEEE Visualization
  2005, Poster Compendium, Minneapolis, MN. October
  2005. (BEST POSTER AWARD)
• Color Rapid Prototyping for Diffusion Tensor
  MRI Visualization. Daniel Acevedo, Song Zhang,
  David H. Laidlaw, and Chris Bull. 7th Int.
  Conference on Medical Image Computing and
  Computer Assisted Intervention. Short paper. St.
  Malo, France, September 2004.
• Graphic Design, Art, and Scientific
  Visualization. Cartoon Future Master.
  Invited talk. European Association of Animation
  Film (Organizers). A Coruña, Spain, April 2004.
• Case Studies in Building Custom Input Devices
  for Virtual Environment Interaction. Joseph
  LaViola, Daniel Keefe, Robert Zeleznik, and
  Daniel Acevedo. VR 2004 Workshop Beyond
  Glove and Wand Based Interaction, Chicago, IL,
  March 2004.
• Designer-critiqued Comparison of 2D Vector
  Visualization Methods A Pilot Study. Cullen
  Jackson, Daniel Acevedo, David H. Laidlaw, Fritz
  Drury, Eileen Vote, Daniel Keefe. SIGGRAPH
  2003 Sketches and Applications Proceedings. San
  Diego, CA. August 2003.
• Design-by-Example A Schema for Designing
  Visualizations Using Examples from Art. Eileen
  Vote, Daniel Acevedo, Cullen Jackson, Jason
  Sobel, David H. Laidlaw. SIGGRAPH 2003
  Sketches and Applications Proceedings. San Diego,
  CA. San Diego, CA. August 2003.
• Discovering Petra Archaeological Analysis in
  VR. Eileen Louise Vote, Daniel Acevedo Feliz,
  David Laidlaw and Martha S. Joukowsky. IEEE
  Computer Graphics and Applications.
  September/October 2002
• Pop Through Buttons for Virtual Environment
  Navigation and Interaction. Robert Zeleznik,
  Joseph LaViola, Daniel Acevedo, and Daniel Keefe.
  Virtual Reality 2002, March 2002.
• Archaeological Data Visualization in VR
  Analysis of Lamp Finds at the Great Temple of
  Petra, a Case Study. Daniel Acevedo, Eileen Vote,
  David Laidlaw and Martha Joukowsky. IEEE
  Visualization 2001. San Diego, California.
  October 2001. (BEST CASE STUDY AWARD)
• CavePainting A Fully Immersive 3D Artistic
  Medium and Interactive Experience. Daniel Keefe,
  Daniel Acevedo Feliz, Tomer Moscovich, David
  Laidlaw and Joseph LaViola. ACM SIGGRAPHs
  I3D 2001 Symposium on Interactive 3D Graphics.
  North Carolina, March 2001.
• Hands-Free Multi-Scale Navigation in Virtual
  Environments. Joseph LaViola, Daniel Acevedo
  Feliz, Daniel Keefe and Robert Zeleznick.
  ACM SIGGRAPHs I3D 2001 Symposium on Interactive
  3D Graphics. North Carolina, March 2001.
• Virtual Reality and Scientific Visualization in
  Archaeological Research. Eileen Vote, Daniel
  Acevedo, Martha S. Joukowsky and David Laidlaw.
  VAST 2000 Virtual Archaeology between
  Scientific Research and Territorial Marketing.
  Arezzo, Italy, November 2000.
• ARCHAVE A Virtual Environment for
  Archaeological Research. Daniel Acevedo, Eileen
  Vote, David H. Laidlaw and Martha S. Joukowsky.
  IEEE Visualization 2000. Work in Progress
  report. Salt Lake City, Utah. October 2000.
• ARCHAVE - A Three Dimensional GIS for a CAVE
  Environment. Eileen Vote, Daniel Acevedo, Martha
  Joukowsky and David Laidlaw. CAA 2000
  Computing Archaeology for Understanding the Past
  Ljubljana, Slovenia April, 2000.