Rendering techniques can improve spatial perception and orientation in 3D virtual environments but may reduce accuracy of point-wise value estimation of thematic data. An experiment evaluated different rendering techniques on tasks of mental mapping, distance estimation, and value estimation using color mappings in a 3D city model. Results showed that edge enhancement and abstracted façade textures improved mental mapping the most, while rendering techniques reduced accuracy for point estimates but increased it for area estimates. The impact of rendering techniques depends on the visualization task.
Impact of Rendering Techniques on 3D Color Mappings
1. Evaluating the Perceptual Impact of Rendering Techniques on
Thematic Color Mappings in 3D Virtual Environments
18th International Workshop on Vision, Modeling and Visualization
Juri Engel Amir Semmo Matthias Trapp Jürgen Döllner
Hasso-Plattner-Institut, Potsdam, Germany
2. 1) Motivation 2) Rendering Techniques 3) User Study
4) Results 5) Conclusions 6) Questions
3. • Using color maps to communicate information is a fundamental approach in visualization
• In 3D virtual environments surface-based data is usually visualized using color-encoded surface textures
Air flow at the surface of a cooling jacket [R.S. Laramee, 2004] Deformation of the colon between two CT scans [W. Zeng, 2004]
4. • Virtual 3D city models communicate geospatial information by means of virtual 3D worlds
• Thematic data is an integral part of geospatial data along geometry, topology, semantics, and appearance
• For example, results of a solar potential, crime, or heat transmission analysis may be mapped to colors
Object semantic Visibility Solar potential
5. • Huge amount of objects
• Objects need to be visually
distinguished
• Most surfaces are flat
• Appearance and
microstructure is usually
modeled with textures
• Appearance is highly
important for the recognition
of landmarks and orientation
6. • 1.8 km² central urban area of
Berlin, Germany
• Solar radiation summed up
over a year
• Considering surface
orientation and shadowing
• Continuous surface
information
• Precomputed and stored in a
texture atlas
• Values: 0-no radiation; 1-
maximum radiation
7. • Objects can hardly be distinguished because of missing monocular depth cues
• Appearance and microstructure information is lost
8. • Depth Cues
• Enable the perception of 2D images as three dimensional
• Essential for mental modeling to estimate distance, relative position, object size and shape
• Monocular depth cues: linear perspective, relative size, texture gradient, occlusion, shading, shadows, areal
perspective, and defocus blur [J. D. Pfautz, 2000]
• Rendering Techniques for Depth-Cue Enhancement
• Edge Enhancement
• Global Illumination
(Ambient Occlusion)
• Local Illumination
(Blinn-Phong Shading)
• Abstracted façade textures
9. • Depth Cues
• Enable the perception of 2D images as three dimensional
• Essential for mental modeling to estimate distance, relative position, object size and shape
• Monocular depth cues: linear perspective, relative size, texture gradient, occlusion, shading, shadows, areal
perspective, and defocus blur [J. D. Pfautz, 2000]
• Rendering Techniques for Depth-Cue Enhancement
• Edge Enhancement
• Global Illumination
(Ambient Occlusion)
• Local Illumination
(Blinn-Phong Shading)
• Abstracted façade textures
10. • Depth Cues
• Enable the perception of 2D images as three dimensional
• Essential for mental modeling to estimate distance, relative position, object size and shape
• Monocular depth cues: linear perspective, relative size, texture gradient, occlusion, shading, shadows, areal
perspective, and defocus blur [J. D. Pfautz, 2000]
• Rendering Techniques for Depth-Cue Enhancement
• Edge Enhancement
• Global Illumination
(Ambient Occlusion)
• Local Illumination
(Blinn-Phong Shading)
• Abstracted façade textures
11. • Depth Cues
• Enable the perception of 2D images as three dimensional
• Essential for mental modeling to estimate distance, relative position, object size and shape
• Monocular depth cues: linear perspective, relative size, texture gradient, occlusion, shading, shadows, areal
perspective, and defocus blur [J. D. Pfautz, 2000]
• Rendering Techniques for Depth-Cue Enhancement
• Edge Enhancement
• Global Illumination
(Ambient Occlusion)
• Local Illumination
(Blinn-Phong Shading)
• Abstracted façade textures
12. • Color information is altered, which may lead to ambiguity in the interpretation of a color mapping
• Each rendering technique has a different impact
• Goal for visualization design: reduce information ambiguity while still preserving depth cues
13. • To which degree improve rendering techniques the orientation in 3D and creation of mental maps?
• Do they improve the perception and estimation of distances?
• Do they alter the perception of the visualized thematic data?
• Perform certain rendering techniques better for thematic visualization than others?
• Does a combination of rendering techniques always improves the orientation and distance estimation?
→ We performed a quantitative and qualitative user study to objectively compare and evaluate the
different rendering techniques
→ The purpose of this study was to determine if there is a significant main effect on the rendering
techniques and color mappings
16. 1. All rendering techniques would improve the spatial perception and, thus, would reduce task time
and error rate for mental mapping and distance estimation.
2. The estimation of thematic values would be more difficult with any rendering technique and would
result in higher error rates and task completion times.
3. There would be significant main effects between color mappings and the tasks’ completion times and
error rates.
4. There would be a distinct order of rendering techniques and color mappings for each task.
5. Compared to individual rendering techniques, a combination would improve the participants’
performance in task 1 and 2 but increase the error rate in task 3 and 4.
17. • Evaluated different color maps to have more generalized results
• Single hue: only saturation as visual variable
• Black body radiation (BBR): hue, saturation, and luminance as visual variables
(Luminance can interfere with the output of rendering techniques)
• Diverging: saturation as variable and two hues that can be easily correlated with high and low value, neutral midpoint
19. • Within subjects design 4×3×8
(task × color mapping ×
rendering technique) = 96 trials
• Additional trial in the
beginning for practicing
• Task and trial order were
randomized to avoid sequence
effects
• Questionnaire for usefulness of
each rendering technique after
completion of each task
• 21 participants (17 male, age
21-49, 19 from campus)
• Average completion time:
40 minutes
20. • Each rendering technique reduced the time required for
orientation over a plain color mapping
• Using abstracted façade textures resulted in the best
performance increase
Rendering techniques: (0) none, (1) Blinn-Phong shading, (2) SSAO, (3) edge enhancement (EE), (4) abstracted facade textures (AFT), (5) Blinn-Phong shading + SSAO, (6) SSAO + AFT, (7) AFT + EE
21. • Applying the examined rendering techniques didn’t result in a
performance improvement
• Participants had difficulties to overview all marked points
→Abstracted façade textures are prone to create visual clutter
Rendering techniques: (0) none, (1) Blinn-Phong shading, (2) SSAO, (3) edge enhancement (EE), (4) abstracted facade textures (AFT), (5) Blinn-Phong shading + SSAO, (6) SSAO + AFT, (7) AFT + EE
22. • As expected, each rendering technique reduced the accuracy of
a point-wise value estimation
• Local illumination showed to be least suitable for an accurate
point-wise value estimation
Rendering techniques: (0) none, (1) Blinn-Phong shading, (2) SSAO, (3) edge enhancement (EE), (4) abstracted facade textures (AFT), (5) Blinn-Phong shading + SSAO, (6) SSAO + AFT, (7) AFT + EE
23. • Surprisingly, each rendering technique increased the accuracy of
area-wise value estimation
• Edge enhancement showed to be the most suited technique for
an accurate area-wise value estimation
Rendering techniques: (0) none, (1) Blinn-Phong shading, (2) SSAO, (3) edge enhancement (EE), (4) abstracted facade textures (AFT), (5) Blinn-Phong shading + SSAO, (6) SSAO + AFT, (7) AFT + EE
24. • Color mapping has a significant effect on value
estimation in 3D virtual environments
• For the mental mapping task the average
completion time using the BBR was 15.1%
longer than using the other two color mappings
• The combination of different techniques shows
no pattern compared to the individual ones
regarding performance
25. • Participants rating: 0-not useful at all, 7-absolutely useful
• Edge enhancement was perceived as the most helpful rendering technique
• Abstract façade textures were perceived helpful for spatial tasks
• Effect of abstract façade textures was underestimated for value estimation tasks.
26. • Rendering techniques can be used to improve mental mapping and orientation in a thematic visualization
• The examined rendering techniques alter the perception of visualized data
• Whereby they reduce the accuracy of a point-wise value estimation
• Color mappings have a significant effect on the perception of thematic data in 3D virtual environments
27. • Rendering techniques can increase
the accuracy of an area-wise value
estimation
• Attention should be paid to visual
clutter
• There was no clear order of
rendering techniques among all tasks
• A combination of multiple
techniques was not necessarily
better than single techniques
→ The used rendering technique
should be chosen according to the
visualization needs
Rendering Technique Mental
Mapping /
Orientation
Distance
Estimation
Point-Wise
Value
Estimation
Area-wise
Value
Estimation
None – + –
Blinn-Phong Shading –
SSAO
Edge enhancement +
Abstracted façade
textures (AFT)
+ –
Blinn-Phong + SSAO –
SSAO + AFT –
Edge enhancement +
AFT
–
28. • Evaluation with color mappings for different data sets
• Follow-up evaluation of edge enhancement for area-wise value estimations
• View-dependent combination of rendering techniques for depth-cue enhancement (e.g., view-distance
based or focus+context visualization)
• Generalization scheme for color-encoded thematic data to reduce visual clutter
29. Thank You For Your Attention!
• Juri Engel
juri.engel@hpi.uni-potsdam.de
• Amir Semmo
amir.semmo@hpi.uni-potsdam.de
• Computer Graphics Group
Prof. Dr. Jürgen Döllner
hpi3d.de youtube.com/hpicgs @hpi3d