1	Evaluating Space Perception in NPR Immersive Environments Amy Gooch and Peter Willemsen University of Utah
2	Immersive Environments Enable us to test the human visual system (HVS) in environments which would otherwise be difficult to generate However, what kind of effects does the information provided in the immersive environment have on the HVS? Depth, scale, and distance should be properly presented
3	Functional Realism Ferwerda at ACM Campfire on Perceptually Adaptive Graphics Physical Realism Photographic Realism Functional Realism A subset of NPR Communicate viewer invariant information
4	Evaluating functional NPR Does the functional realism of NPR impart a sense of space when used in immersive environments? Previously looked to artists for how to convey Should look to perceptual psychologists on what to convey
5	Space Perception Where? In the world Relative to viewer How big? Scale Motion
6	“Perception of scale and distance is poor in computer graphics” Is this true? (insert picture here) Why is it true? Can we do anything about it?
7	Is it a problem? Games
8	What do we mean by scale and distance? Spatial organization Egocentric vs Exocentric
9	What do we mean by scale and distance? Spatial organization Absolute vs Relative Visual cues from one may not help the other
10	Size and distance are closely related
11	Size and distance are closely related
12	Questions What (spatial) information is potentially available? What (spatial) information in images do viewers use? What can you accurately convey in CG?
13	Perception and action Can’t (shouldn’t) consider visual perception in isolation Biomechanical perception
14	Previous Research Perceptual Psychologists Rieser et al. 1990 (Vanderbilt University) Loomis et al. 1992 (UCSB) Philbeck and Loomis 1997 Witmer and Kline 1998
15	Previous Research
16	Perceptually guided action tasks We want to: Examine perception of absolute egocentric distances in immersive environments Absolute egocentric distance refers to the actual metric distance between an observer and an external object
17	Perceptually guided action tasks How to test for perception of absolute egocentric distance? Verbal reporting subject to bias Directed action tasks: Previous research has shown that people are accurate at judging distances between 2 and 25 meters
18	Perceptually guided action tasks Direct walking task: Subject asked to build a “good image” of environment, concentrating on target With subject blindfold/HMD screen blanked subject walks purposefully to the target location and stop when standing on the target Subject doesn’t receive feedback
19	Perceptually guided action tasks Pretest: Random-dot stereogram eye test Checked for 20/20 vision with Snellen Chart Read instructions Listen and watch experimenter give an example of the task
20	Perceptually guided action tasks Equipment and training: Subjects wear white-noise generating mono-headphones to help remove localization Collar to physically restrict view of feet
21	Perceptually guided action tasks Equipment and training: Practice walking blindfold with collar and headphones (5 minutes) Lead blindfolded into experiment area
22	Perceptually guided action tasks Perceived egocentric distances in real and immersive environments using an action task
23	Perceptually guided action tasks Subjects view environment and target, then walk without vision (blindfolded) to perceived location of target
24	Equipment nVision Datavisor Hi-Res 1280 x 1024 interlaced CRT 52^o Field of View, 100% stereo overlap Rendered on SGI Onxy2 R12k, 2 IR pipes IPD: 6.5cm, M=6.22, [5.25, 7.0] At least 25 fps Intersense IS600-Mark2 Tracker (5’x10’)
25	Apply these techniques to evaluate NPR
26	NPR VR Experiment Distances: 2m, 3.5m, 5m 6 males, 6 females 2 conditions: Real world Feature lines VR 12 trials per condition, including 3 practice trials
27	Results
28	Results
29	Follow-up study
30	Conclusion Our study shows that simple NPR worlds can at least produce distance estimation in the same ball park as shown by previous studies (50-70%)
31	Conclusion Running experiments is hard and must be done correctly to be of value to the computer graphics and psychology community Instructions must be consistent and carefully worded Collaborate with psychologists Help determine number of subjects, number of trials, etc.
32	Looking forward... What does it take to make an NPR environment communicate as much as a traditionally rendered environment?
33	Thanks to…. Pete Willemsen & Bill Thompson Sarah Creem-Regehr, John Rieser, Herb Pick, Jack Loomis for attempting to teach me about perceptual psychology Bruce Gooch Alias\|Wavefront for Maya NSF CDA-9623614 and 58500700
34	Announcement Second Annual SIGGRAPH NPR BOF When: Sunday, July 21, 2002 Where: Patio Room at the Menger Hotel, San Antonio, Texas Time: 8:30pm --- See http://www.cs.utah.edu/npr for details
35	"Cue" Cue Accomodation Binocular convergence Binocular disparity Familiar size Relative size Linear perspective Height in picture
36	"Cue" Cue Ariel perspective Motion parallax Relative motion Texture gradients Shading Occlusion
37
38	Perceiving Depth Ocular Information accomodation, convergence Stereoscopic Information Dynamic Information Motion parallax, moving observer/object, accretion/deletion of texture Pictorial Information linear perspective, position relative to horizon, relative size, familiar size, texture gradients, edge interpolation, shading, aerial perspective...
39	Perceiving Depth Stereoscopic Information (add picture)
40	Perceiving Depth: Dynamic Information Motion Parallax (add picture)
41	Perceiving Depth: Dynamic Information Moving Observer (add picture) Moving Object (add picture) Accretion/Deletion of Texture (add picture)
42	Perceiving Depth: Ocular Information Accomodation (Add pictorial example, how it applies to immersive environments)
43	Perceiving Depth: Ocular Information Convergence (Add pictorial example, how affects HMD)
44	Perceiving Depth: Pictorial Information Linear Perspective
45	Perceiving Depth: Pictorial Information Convergence of Parallel Lines
46	Perceiving Depth: Pictorial Information Position Relative to Horizon on Surface (add picture)
47	Perceiving Depth: Pictorial Information Relative Size (add picture)
48	Perceiving Depth: Pictorial Information Familiar Size (add picture)
49	Perceiving Depth: Pictorial Information Texture Gradients (add picture)
50	Perceiving Depth: Pictorial Information Edge interpretation (add picture)
51	Perceiving Depth: Pictorial Information Shading Information
52	Perceiving Depth: Pictorial Information Aerial Perpective (add picture)
53	Defining Terminology Immersive Environments Functional NPR Space and Depth Perception