Stereoscopic depth from absolute and relative disparities
Adrien Chopin David Knill Dennis Levi Daphne Bavelier
or
The phenomenon of absolute depth "blindness"
VSS 2014
Corresponding author:
[email protected]; download at http://adrien-chopin.weebly.com/publications.html
+ φL
φR
a P = φL - φR
φL
φL
-
Fixation
P2 φL
φR
φR
φR r = φL - φR or r = a P1 - a P2
Are absolute and relative depth systems independent? It is an unresolved question: are relative disparities independently acquired or are they 1 computed as a difference of absolute disparities ? Some believe that the two systems are independent because relative depth performance is better than absolute depth performance. However several concerns need to be addressed. Concern 1: The absolute depth condition always presents a visible reference (fixation or screen border), creating relative disparities.2
Control group Dressmakers
2000
1000
0
Fixation
+
3000
Absolute Depth
Relative Depth
Stereoblindness and absence of fine stereopsis are more likely for absolute depths than for relative depths. 86% of the control participants lack fine stereopsis for absolute depths. Absolute depth training did not yield benefits after 5 sessions. Completely stereoblind 0.8
Control group Dressmakers
0.6 0.4 0.2 0
N = 21 N = 13
R lute elative D Dept epth h
Abso
Absolute depth training
No fine stereo (>900’’) 0.8 0.6 0.4 0.2 0
3000
1000 N=6 0
0
1
2
3
4
5
Sessions #
R lute elative D Dept epth h
Abso
Concern 2: Absolute depth performance is typically measured with 2IFC and relative depth with 2AFC giving the advantage to relative depths (because of memory decay).3
Solution: Participants were naive: either students or professional dressmakers, who may particularly benefit from a good stereopsis in their activity.
Solution: We used a single stimulus method, introducing the same memory burden in the two conditions. Relative Depth Absolute Depth
Result: Dressmakers are more precise than the control group (p