Does the Quality of the Computer Graphics Matter When
Judging Distances in Visually Immersive Environments?
Authors: Thompson, Creem-Regehr, et al.
Presenter: Jeff Peirson
2 of 29
The Problem
• In visually immersive environments, distances are significantly compressed
• This affects a user’s ability to judge distances and effectively interact with the virtual environment
3 of 29
The Question
• Previous Studies: What effect does…– limited FOV have?– difficulties in accurately using binocular stereo
in HMDs have?– limited resolution have?
• This Study:– What effect does image quality have in distance
judgments?
4 of 29
Presentation Outline
• Useful background information
• Explanation of the research experiment
• Results & analysis of findings
• Reasoning behind results, & discussion
5 of 29
Background: Terms
• Visual perception terms:– Egocentric distance:
distance measured from observer to an object
– Exocentric distance: distance measured between two objects
6 of 29
Background: Terms
• Visual perception terms:– Absolute distance: “Point A is 5 meters away”– Relative distance: “Point A is twice as far
away as point B”– Ordinal distance: “Point A is behind point B”
7 of 29
Background: Terms
• Visual perception terms:– Personal space: 0m – 1m– Action space: 1m – 30m– Vista space: > 30m
8 of 29
Research of Interest• This research:
– absolute egocentric distance judgments in action space
9 of 29
Background: Common Visual Cues
• The effect of some common visual cues:
10 of 29
Previous Research
• Critical problem: How do we obtain unbiased distance judgments?
• (A) Verbal reports (explicit reporting)– difficult for subject to accurately report
distance– difficult to control biases, such as anchoring
• (B) Perform an action (implicit reporting)– have subject implicitly determine distance
11 of 29
Previous Research
• Critical problem: How do we obtain unbiased distance judgments?
• Perform an action… how?– (A) walk directly towards target blindfolded, and
stop when target is presumably reached (direct walking)
– (B) walk in oblique direction towards target, then move a couple steps directly towards target (triangulated walking)
12 of 29
Previous Research
• Triangulated Walking… what is it again?
13 of 29
• Previous findings…
• Found using virtual environments similar to…
Previous Research
14 of 29
Authors’ Experiment:The Conditions
• Test a group of subjects under any one of four conditions:– Real-World– High-Resolution Panoramic Image VR– Low-Quality, Texture-Mapped VR– Wireframe VR
15 of 29
Authors’ Experiment:The Subjects
• 48 college students– Four groups of 12 (one group for each condition)– Each group was 50/50 male/female
• All subjects had normal or corrected-to-normal vision
• Interpupilar distance actually ranged from 5.1cm to 7.7cm!
16 of 29
Authors’ Experiment:The Apparatus
• The target: A ~14-inch foam disk– placed at either 5m, 10m, or 15m from subject
• nVision Datavisor HiRes HMD– 1280x1024 res., 42º FOV
• InterSense IS600 Mark 2 tracker– used only head rotation to update viewpoint
17 of 29
Authors’ Experiment:The Conditions, Explained
• Real-World:– An engineering building lobby
• Three Virtual Worlds of Differing Quality:– A rendering of the same lobby– Target placed in same locations as real-world condition
18 of 29
Authors’ Experiment:The Conditions, Explained
• 1) High-Res., Realistic, Panorama– Essentially mapped sets of camera images to
make a 360º virtual world– Multiple image sets were made at 5cm height
intervals to accommodate subject eye height• If you were an inch taller or shorter, would you
perceive distances to the ground differently?
– No motion parallax cues– “Slightly incorrect stereo geometry”
19 of 29
Authors’ Experiment:The Conditions, Explained
• 2) Low-Res. Texture-Mapped– Same geometry as lobby– Simple, tiled texture maps– Simple, point-source lighting– No shadows, no global illumination effects
20 of 29
Authors’ Experiment:The Conditions, Explained
• 3) Wireframe– Same as before, except an OpenGL silhouette
algorithm was run over the virtual environment– Result is a black on white “sketch”
21 of 29
Authors’ Experiment:The Procedure
• Step I) Training1. Subject given a written set
of instructions detailing triangulated walking
2. Subject shown a demonstration in a mini-setting
22 of 29
Authors’ Experiment:The Procedure
• Step I) Training3. Subject told to get a “good image” of the target
in the environment to accustom subject to his/her surroundings (either real or virtual)
– Subject was not allowed to move his/her head from side to side
4. Practice triangulated walking
23 of 29
Authors’ Experiment:The Procedure
• Step I) Training– Eye height may be changed if subject is
allowed to look directly at their feet– Solution: the subject is never allowed to see
their feet
24 of 29
Authors’ Experiment:The Procedure
• Step II) Transport the subject– Transport the subject to the testing area blind and
led by an instructor– This keeps the subject from building a mental
model of the room and walls beforehand, especially for VR conditions
25 of 29
Authors’ Experiment:The Procedure
• Step III) The actual experiment– Subject wears a microphone and is fed
instructions for triangulated walking
26 of 29
ResultsThe result: Image quality
has very little effect over compression effect in immersive environments
Avg. ratio of perceived dist. to actual dist.:
• Real-world: ~100%• High-Res CG: 50%• Low-Res CG: 44%• Wireframe CG: 40%
27 of 29
Reflections & Discussion
• High-res. environment was extremely realistic, yet it did not make much of a difference– Why???
• Possible solution to compressed perception:– Allow subject to freely explore environment– This effectively allows the subject to become
accustomed to the compressed virtual world– But… is this useful?
28 of 29
Reflections & Discussion
• In conclusion: If image quality is not a critical factor in distance judgment, then what is?– Limited sense of presence?– Limited FOV?– Lack of motion parallax?– Ergonomic effects of HMD?
29 of 29
Questions? Ideas?