CHI 97 Electronic Publications: Technical Notes
How Effective are 3D Display Modes?
Sabine Volbracht
Dept. of Computer Science
University of Paderborn
Fürstenallee 11
D-33102 Paderborn
sabaro@uni-paderborn.de
Gitta Domik
Dept. of Computer Science
University of Paderborn
Fürstenallee 11
D-33102 Paderborn
domik@uni-paderborn.de
Khatoun Shahrbabaki
SAP AG
Postfach 1461
D-69185 Walldorf
khatoun.shahrbabaki@sap-ag.de
Gregor Fels
Dept. of Organic Chemistry
University of Paderborn
Fürstenallee 11
D-33102 Paderborn
gf@chemie.uni-paderborn.de
ABSTRACT
The increasing availability of 3D input and output
devices demands a better understanding and comparison of their
quality. This report describes an empirical experiment for comparing
3D display modes: traditional perspective viewing, anaglyph stereo and
shutter glass stereo. We followed two hypotheses 1. shutter glass
stereo viewing allows a faster and more accurate recognition than the
anaglyph and the perspective viewing, and 2. subjects experienced with
particular 3D representations are faster and more accurate than
subjects with out experience. The experiment is based on a true
research scenario in organic chemistry. Organic molecules were used as
3D objects. Mean response error and mean response time were calculated
for a series of six tasks and 81 subjects.
KEYWORDS
perspective viewing, anaglyph stereo, shutter glass stereo, 3D
display mode, experience, empirical, experiment.
© Copyright ACM 1997
INTRODUCTION
Former developments in computer technology have led to an
increasing interest of stereo techniques. Consequently it is an
interesting issue if stereo effects provide more expressiveness
and effectiveness in comparison to the traditional display techniques and what kind of display mode is adequate for a particular user.
Previous studies in this area have centered around the comparison of the perspective viewing mode versus the shutter glass
stereo mode. They indicate that shutter display mode provides
better user performance at many 3D visual tasks than perspective mode does [1][4]. Our study differs in that it reduces the
testing environment to one application area (namely chemistry), one visualization technique (the molecule stick model)
and three 3D display modes (perspective viewing, anaglyph
stereo and shutter glass stereo with StereoGraphics Crystal
Eyes" LCD shutter glasses). This set-up provides a controlled
testing environment resulting in quantitative comparisons,
which are meaningful for the decision-making of a potential
buyer. The visualization technique as well as the tasks used in
our experiment are taken from a true research scenario in
organic chemistry. Structures are represented as stick models,
easily understood by the organic chemist. Stick models are
used to show the structure of the molecule.
EXPERIMENTAL DESIGN
The basic design [2] was a 3x3 factorial experiment with three
classes of the independent variable 'display mode' ( perspective
viewing, anaglyph stereo and shutter glass stereo ) and three
levels of the independent variable 'experience' ( high, low, none
experience ). The dependent variables were 'accuracy' and
'time'.
SUBJECTS
Our experiment involved 81 participants. The subjects were
students of Chemistry (specialized in organic / other) or Computer
Science and had different expertise with 3D representation
of organic molecules.
TASKS
Three problems (identifying, comparing and positioning) were
tested by providing five tasks. Identifying and comparing were
each tested with a simple and a complex molecule to under
stand the relationship between complexity and viewing. The
five tasks were as follows:
- T1 and T2. Identifying (simple and complex molecule)
Count the rings in the molecule.
- T3. Comparing (simple molecule)
Which atom is the nearest and which one the most distant
from the viewer?
- T4. Comparing (complex molecule)
Determine the order of the benzene rings on the z-axis.
- T5. Positioning
Position the benzene ring parallel to the screen plane.
PROCEDURE
For each task the subject saw a different molecule. In this way
we avoided subjects remembering the structure of molecules.
A counterbalancing procedure provided a between-subject
design for each task. The tasks were performed interactively on
each of the 3D display modes. The interaction was restricted to
molecule rotations with the mouse. An objective comparison
was based on the correct answers of the questions and the measured time for answering.
Before beginning the experiment and before each task, the subjects received practice time for a few minutes to become familiar
with each 3D display mode and with the particular
molecule. Once the subject pressed a 'Start/Ready' button, the
interviewer explains the task. After performing the task, the
subject pressed the 'Start/Ready' button again to indicate the
end of the trial. At that time the response time and other necessary
information were recorded.
RESULTS AND DISCUSSION
Mean response errors and mean response times were computed
separately by a two-way Analysis of Variance (ANOVA) with
9 experimental conditions. After that a Newman-Keuls test
( a = 0.05) was applied for comparing the different mean times
and errors. In both cases (response errors and times) interaction
effects were not significant. Results in the following four figures
are averaged over 27 subjects and 3 experience levels.
|
| Figure 1: Frequencies of answers by display mode for P
(perspective) and A (anaglyph) and S (shutter glasses) on
response errors . The underlined numbers represent the
correct answers.
|
|
| Figure 2: Main effect of display
mode for P (perspective) and A
(anaglyph) and S (shutter glasses)
on response errors .
|
|
| Figure 3: Main effect of display
mode for P (perspective) and A
(anaglyph) and S (shutter glasses)
on response errors .
|
|
| Figure 4: Main effect of display mode for P (perspective), A (anaglyph) and S (shutter glasses) on response time .
|
In the following two tables the degrees of freedom associated
with all F-ratio are 2 and 72. So we will replace the usual notation 'F(2,72) =' by 'F = '.
|
| Table 1: Summary of computed F-ratios of display mode.
|
|
| Table 2: Summary of computed F-ratios of experience level .
|
|
| Figure 5: Summary of correct answers from T1 to T5. 100% is correlated with 135 (= 5 x 27) correct answers with a low tolerance of errors.
|
|
| Figure 6: Mean response time for the correct answers of T1 to T5.
|
A more detailed description of the experiment and the results
can be found in [3].
For T2, where the molecule was more complex than for T1, a
Newman-Keuls test indicated, that identifying in shutter and
anaglyph mode has been performed more accurate and faster
than in perspective mode.
In tasks three and four (T3, T4), where the spatial information
was relevant, viewing in perspective mode was considerably
worse than in stereo modes, as expected. Only for T4 the differences of the shutter and the anaglyph mode are also significant. The analysis of the obtained data from T5 showed that
positioning errors were smaller in anaglyph and shutter mode
than in perspective mode but the position time was only significantly better in anaglyph mode. The difference in mean time of
shutter and perspective mode was not significant.
CONCLUSIONS
A very interesting outcome was that viewing in the anaglyph
mode shows a strong resemblance to the quality of shutter
mode. A comparison of cost vs. performance of the three here
discussed 3D display modes would therefore favor anaglyph
stereo. The reader should note that the main disadvantage of
anaglyph stereo, namely the lack of color attributes, is not considered in this comparison.
Our experiment also demonstrated that user experience is relevant for identification of special objects. No indication was
given that level of experience with special objects changed the
results on comparing or positioning tasks. We expect similar
results for visualization techniques in other application areas,
e.g. flow charts or networks.
ACKNOWLEDGMENTS
We thank all participants of our experiment. This research was
partly sponsored by the Ministerium für Wissenschaft und Forschung in Nordrhein-Westfalen and SAP AG.
REFERENCES
|
1. R. J. Beaton and N. Weiman: User evaluation of cursor
positioning devices for 3D display workstations. Three-Dimensional Imaging and Remote Sensing Imaging, SPIE
Proc. 902, 1988, pp. 53-58.
|
|
2.Ray E. Eberts: User Interface Design. Prentice-Hall, Inc.,
1994.
|
|
3.Sabine Volbracht, Gitta Domik, Khatoun Shahrbabaki,
Gregor Fels: An Experimental Comparison of 3D Display
Modes. Proceedings of IEEE Visualization 1996, Late Breaking Hot Topics Papers, pp. 8-11, San Francisco, CA.,
USA.
|
|
4.Colin Ware and Glenn Franck: Evaluating Stereo and
Motion Cues for Visualizing Information Nets in Three
Dimensions. ACM Transactions on Graphics , Vol. 15, No.
2, April 1996, pp. 121-140.
|
CHI 97 Electronic Publications: Technical Notes
