Transcript PPT - Andrew T. Duchowski
Eye Tracking With Stereoscopic Images
Eamon Moore, Punit Seth, Dhaval Shah Clemson University
Introduction
Stereoscopic image – optical illusion of depth seen by focusing ones eyes in front of or behind an image [7] Each eye views an image differently which gives the perception of depth.
Eye Tracking
Eye Trackers – Can be used to track eye movements and gaze coordinates Gaze coordinates – Helps in understanding why some people see stereo images and some do not
Divergence and Convergence
Divergence and Convergence – the methods that people use to view stereograms Divergence – Moving your eyes outward in the opposite direction Convergence – Moving your eyes inward
Why Use Stereograms?
Marketers and researchers – Attempts are being made to utilize ones ability to see three-dimensional images and use them in advertising.
Stereograms can enhance vividness, clarity, realism, and depth.
The Experiment
Analyzing the behavior of the eyes to view stereograms [dependant variable] Convergence Divergence Looking for significant differences in Placebo and Experimental group [independent variable]
Hypothesis
Null Hypothesis – There will be no significant change in the distance of the eyes when viewing stereograms, regardless of experimental condition.
Alternate Hypothesis – There will be significant results that indicate divergence of the eyes in both conditions.
Background
Brain processing – The brain accepts two images that are seen by each eye and creates a completely different three dimensional picture called stereo [6].
Figure 1: Image processing Stereo allows you to see objects as solids in dimension of width, height, and depth.
When Stereoscopy Started
Idea of stereoscopy preceded photography Paintings were made by Giovanni Porta in the late 1500s by placing images side by side. This showed his understanding of binocular vision.
Three-Dimensional Glasses
Three-Dimensional Glasses – red filter for left eye, blue filter for right eye [11] Figure 2: Red-blue Stereo Image When looked at images that have depth, a three-dimensional image could be seen.
Modern Stereogram
First modern stereogram created in 1959 by Julesz [11] Original image viewed by left eye Modified version of original image viewed by right eye Brain fuses both images creating the final image Figure 3: Modern Stereogram
Single Image Stereogram
Created in 1979 by a student of Julesz, Tyler Found that the offset idea could be applied to a single image to create a black and white random dot stereogram Figure 4: Single Image stereogram
Colored Stereogram Program
In 1991 Smith improved on the research of Julesz by creating stereogram modeling software.
Eliminated the need for dots and provided color
Tracking of Eye Movements and Visual Attention
Study conducted by Neuroinformatics Group, Bielefield University [8] Concentrated on vergence eye movements using stereograms similar to the ones used in this experiment Figure 5: Coarse Granularity Image (left) ; Stereogram (right)
Neuroscience Institute
Gave insight about vergence eye movements Discussed dynamics of horizontal and vertical vergence Study indicated that horizontal eye movements were of more importance.
Program to Create Stereograms
School of Electrical and Electronic Engineering at the University of Nottingham [3] Created program that produces stereograms Examined how stereograms were viewed
Experimental Design
Apparatus
Tobii Eye Tracker [16] – Video-based combined pupil and corneal reflection eye tracker 2.4 GHz 256 MB RAM Windows XP Red Hat Linux Release 9, Version 2.4.20
Sampling Rate = 50 Hz Accuracy = 1º visual angle Figure 6: Tobii System
Experimental Design
Between subjects Two conditions : Experimental group – Stereogram Placebo group – Nonstereo Image 10 Participants
Stimulus - Control Image
Stimulus - Stereogram
The Hidden Image
Stimulus – Nonstereo Image
Salient Features
Reduced calibration points An organized file structure Validity = 0 Timer Shortcut keys Analysis option
Algorithm
Record X L , X R , Y L , Y R .
Distance = Control distance Experimental distance
Algorithm
If (Experimental distance < Control distance) If (X L < X R )
Convergence
Else
Convergence with crossover
. else If (Experimental distance > Control distance)
Divergence
else
No difference
.
Data Analysis
Data Analysis – Experimental Group (Individual)
500
Individual Experimental Moving Averages Compared to Aggregate Control Moving Average
-1000 -1500 0 0 -500 20000 40000 60000 80000 100000 120000 Experimental Subject 1 Experimental Subject 2 Experimental Subject 3 Experimental Subject 4 Experimental Subject 5 Control
Time (milliseconds)
400 200 0 0 -200 -400 -600 -800
Data Analysis – Placebo Group (Individual)
Individual Placebo Moving Averages Compared to Aggregate Control Moving Average
20000 40000 60000 80000 100000 Placebo Subject 1 Placebo Subject 2 Placebo Subject 3 Placebo Subject 4 Placebo Subject 5 Control 120000
Time (milliseconds)
200 0 -200 -400 -600 -800 -1000 1000
Data Analysis – Experimental Group (Aggregate)
Aggregate Experimental Values Compared to Aggregate Control Values
800 600 Experimental Trendline - Polynomial (6th degree) Experimental Trendline - Polynomial (6th degree) Treatment Trendline - Moving Average (Every 255 pts) 400
20000 40000 60000 80000 100000 120000 Time (Milliseconds)
500 400 300 200 100 0 1 -100 -200 -300 -400 -500
Data Analysis – Placebo Group (Aggregate)
Aggregate Placebo Values Aggregate Compared to Control Values
Placebo Trendline - Polynomial (6th degree) Placebo Trendline - Moving Average (Every 255 pts) Control Trendline - Polynomial (6th degree)
Time (Milliseconds)
One Way Analysis of Variance (ANOVA)
Assumptions of an ANOVA Independence Homogeneity of Variance Normality Levene Statistic 3.335
100 0 -100 -200 -300 -400 -500 -600 -700 1 2 3 Case Number df1 1 4 df2 8 Sig.
.105
5 6 7 8 9 10
Descriptive Statistics
Randomly Assigned Groups Placebo - Five Men Experimental - Three Men, Two Women Distance N 10 Min -599.35
Max.
33.28
Mean Statistic -133.4980
Mean Std. Error 60.4466
Std. Dev.
191.1490
Variance 36537.957
ANOVA
• Not a significant difference between the Placebo (M = -36.048, S = 86.891) and Experimental Group (M = -230.949,S = 225.562) Between Groups Within Groups Total Sum of Squares 94965.981
df 1 328841.610
9 Mean Square 94965.981
233875.629
8 29234.454
F 3.248
Sig.
.109
100 0 -100 -200 -300 -400 -500 -600 -700 N = COND 5 1.00
6 5 2.00
Placebo
ANOVA and Power Analysis
N Mean 5 -36.0475
Std. Deviation 86.8910
Std. Error 38.8589
95% Confidence Interval for Mean Lower Bound Upper Bound -143.9370
71.8419
Minimum -163.99
Maximum 33.28
Experimental Total 5 -230.9485
225.6521
10 -133.4980
191.1490
100.9147
60.4466
-511.1326
-270.2378
49.2355
3.2418
-599.35
-599.35
-48.72
33.28
• Post Hoc G-Power Analysis -power of .1077 indicates approximately 11 percent chance that the null hypothesis could have been rejected.
Discussion
Discussion
• Stereograms are viewed by using convergence regardless of stimuli.
• No significant results • Experimental group shows trend towards divergence near the end.
• Placebo group shows a lesser trend towards convergence
Experimental Group (Aggregate)
Aggregate Experimental Values Compared to Aggregate Control Values
1000 800 600 -400 -600 -800 -1000 400 200 0 -200 Experimental Trendline - Polynomial (6th degree) Experimental Trendline - Polynomial (6th degree) Treatment Trendline - Moving Average (Every 255 pts)
20000 40000 60000 80000 100000 120000 Time (Milliseconds)
500 400 300 200 -200 -300 -400 -500 100 0 1 -100
Placebo Group (Aggregate)
Aggregate Placebo Values Aggregate Compared to Control Values
Placebo Trendline - Polynomial (6th degree) Placebo Trendline - Moving Average (Every 255 pts) Control Trendline - Polynomial (6th degree)
Time (Milliseconds)
Limitations
• Low Power - Priori Power Analysis • Tobii Eye Tracker • Stereograms are harder to view on a computer screen.
Future Work
Larger sample size Introduce Z coordinate for the distance from the screen Measure characteristics such as the diameter of the pupil while studying its behavior.
Conclusion
Our hypothesis was incorrect; however, we were correct in believing both groups would behave similarly.
Stereograms are viewed by converging ones eyes; however, a higher power study may prove otherwise. More research can now be conducted to understand how stereograms can be used for advertising, marketing, and other practical applications.
Acknowledgements
Dr. Andrew Duchowski, PhD.,
Associate Professor, Clemson University .
Ms. Puja Seth, M.A.
Doctoral Student, University of Georgia
Mr. Jacob Hicks
Undergraduate Student, Clemson University.
References
[1] Academy of Marketing Science Review. Three-Dimensional Stereographic Visual displays in Marketing and Consumer Research. Available at: http://www.vancouver.wsu.edu/amsrev/theory/holbrook11- 97t.htm. Last Accessed: 10 October, 2004. [2] Annals of the New York Academy of Sciences 2002. Binocular Eye Movement Responses to Dichoptically Presented Horizontal and/or Vertical Stimulus Steps. Available at: http://www.annalsnyas.org/cgi/content/full/956/1/487. Last Accessed: December 2, 2004.
[3] BBC, Nottingham. SIRDS: An optical illusion. Available at: http://www.bbc.co.uk/nottingham/features/2003/08/sirds.shtml#what Last Accessed: December 2, 2004.
[4] CIT,Cornell University. How To See A Magic Eye Poster.
Available At: http://instruct1.cit.cornell.edu/courses/psych470/To_Be_Edited/How%20To%20See%20A%20Magic%20Eye%20P oster%20(MVW).doc. Last Accessed: December 2, 2004. [5] C. Rashbass & G. Westheimer J. Physiol. Disjunctive Eye Movements. 159, 339-360, 1961 [6] Cooper, Rachel. What is Stereo Vision?. 2004. Available At: http://www.vision3d.com/stereo.html. Last Accessed:16 September 2004. [7] Dictionary.com. Available at: http://dictionary.reference.com/search?q=stereogram Last Accessed: December 2, 2004.
[8] Essig, Kai and Ritter, Helg. Tracking of Eye Movements and Visual Attention. Available at: http://www.techfak.uni-bielefeld.de/ags/ni/projects/eyetrack/eye_autostereo.html. The Neuroinformatics Group. Bielefeld University. Last Accessed: 10 October, 2004.
References
[9] Faul, F., & Erdfelder, E. (1992). G-Power: A priori, post- hoc, and compromise power analyses for MS-DOS (computer program). Bonn, FRG:Bonn University, Department of Psychology.
[10] History of Photography and the Camera. Available At: http://inventors.about.com/library/inventors/blphotography.htm
Last Accessed: December 2, 2004.
[11] Magic Eye Inc®. Frequently Asked Questions. 2004. Available at: http://magiceye.com/faq.htm. Last Accessed: 16 September 2004. [12] Mowforth, P. et al. Vergence Eye Movements Made in Response to Spatial-Frequency-Filtered Random-Dot Stereograms. Perception, 10, 299-304, 1981 [13] Patrick Hahn. The History of Stereograms 1996. Available At:http://www2.vo.lu/homepages/phahn/rds/history.htm. Last Accessed: December 2, 2004.
[14] Robert Leggat. Stereoscopic photography 2003. Available At: http://www.rleggat.com/photohistory/history/stereosc.htm.
Last Accessed: December 2, 2004. [15] Sandin, Daniel et al. The VarrierTM Auto-Stereographic Display. Available at http://www.evl.uic.edu/todd/varrier/VarrierSPIE.html. Electronic Visualization Laboratory. University of Illinois at Chicago. Last Accessed: 10 October, 2004. [16] Tobii Technology. User Manual. Available at : http://andrewd.ces.clemson.edu/courses/cpsc412/docs/UsersManual_TobiiClearView_2_1_0.pdf
Last Accessed: December 2, 2004.