Predicting Visual Depth Judgment from Multiple Discrepant Sources of Depth
Information Produced by a Computer Controlled Stereoscopic Camera System

ABSTRACT
Three-dimensional scenes and especially those produced by remote viewing systems
contain discrepant sources of depth information.  The purpose of the study is to
compare two models that predict how people combine discrepant sources of depth
information into a single depth judgment.  The weak fusion model uses the mean of the
depth information sources, while the modified weak fusion model constrains depth
information from the different sources to predict depth judgments. The predictions of
the two models are compared to the depth judgments of subjects viewing discrepant
sources of depth information.  

Subjects in the age range from 18 to 32 years were drawn from University of Hawai’i
psychology classes.  Forty-one of 59 subjects met the criteria for their data to be used
which were: 1) no previous experience with remote three dimensional video viewing
systems, 2) at least 20/40 corrected vision in each eye, 3) ability to perceive two levels
of depth on a stereoscopic test, 4) completion of the experiment and 5) a direct view
score within two standard deviations of the group mean. 

A computer controlled stereoscopic camera system and device for making rapid
changes in target characteristics was developed to produce the discrepant sources of
depth information.  Subjects judged the separation between two point source lights
located at four distances from each other.  Three levels of camera convergence angle,
camera separation and target intensity, along with four different separations between a
pair of lights created 108 viewing conditions within a single testing session.  Linear
regression results indicated that the predictions of the weak fusion model accounted for
more of the depth judgment variance than the modified weak fusion model. The results
are discussed in reference to situations where the weak fusion model can be applied.