Electrical Engineering
      and Computer Sciences

Electrical Engineering and Computer Sciences

COLLEGE OF ENGINEERING

UC Berkeley

PSYCHOPHYSICS & COMPUTATIONAL MODELING OF VISUAL MOTION PERCEPTION

Siddharth Jain

EECS Department
University of California, Berkeley
Technical Report No. UCB/EECS-2007-97
August 7, 2007

http://www.eecs.berkeley.edu/Pubs/TechRpts/2007/EECS-2007-97.pdf

The goal of the research described in this dissertation is to understand the mechanisms by which the brain senses motion. I have performed a detailed psychophysical characterisation of visual motion perception in general and the peculiar omega effect originally discovered by Rose & Blake in particular in which dynamic random noise in the form of random dots displayed in a circular annulus evokes the illusion of rotary motion. I have also found that a model based on the Watson & Ahumada motion detector is able to explain most and key parts of the psychophysical data such as the very delicate effects of frame duration on motion perception, independence of observer performance on dot density in the display and the surprising reverse phi motion caused by contrast reversing dots. In addition to explaining the psychophysical data, the model relates reasonably well to what is known about the neurobiology of motion sensitive cells in the brain making it a realistic model of human visual motion sensing. Some other highlights of the dissertation are as follows: ¿ I find that the intrinsic cortical noise in the brain which manifests itself as uncertainty in motion estimation can play an important role in perception by significantly improving detectability of subliminal motion cues at the expense of a very modest drop in performance for a suprathreshold signal ala stochastic resonance. ¿ I also did experiments on observers under the influence of marijuana and found that the THC in marijuana can cause an impairment of motion perception abilities ¿ observer performance decreases by as much as 15% and reaction time increases by as much as 222±96 ms. ¿ I find that the observer performance is invariant to dot density in the display and argue that this provides very powerful evidence against motion models based on matching dots to nearest neighbors in successive frames ala (Ullman, 1979; Dawson, 1991) etc. ¿ I find and prove that the rotary motion signal does not depend on the center of rotation relative to which it is computed which explains the experimentally observed position invariance of MST(d) cells found by (Graziano, Andersen, & Snowden, 1994).

Advisor: Donald A. Glaser and William J. Welch


BibTeX citation:

@phdthesis{Jain:EECS-2007-97,
    Author = {Jain, Siddharth},
    Title = {PSYCHOPHYSICS & COMPUTATIONAL MODELING OF VISUAL MOTION PERCEPTION},
    School = {EECS Department, University of California, Berkeley},
    Year = {2007},
    Month = {Aug},
    URL = {http://www.eecs.berkeley.edu/Pubs/TechRpts/2007/EECS-2007-97.html},
    Number = {UCB/EECS-2007-97},
    Abstract = {The goal of the research described in this dissertation is to understand the mechanisms by
which the brain senses motion. I have performed a detailed psychophysical characterisation
of visual motion perception in general and the peculiar omega effect originally discovered
by Rose & Blake in particular in which dynamic random noise in the form of random dots
displayed in a circular annulus evokes the illusion of rotary motion. I have also found that
a model based on the Watson & Ahumada motion detector is able to explain most and
key parts of the psychophysical data such as the very delicate effects of frame duration
on motion perception, independence of observer performance on dot density in the display
and the surprising reverse phi motion caused by contrast reversing dots. In addition to 
explaining the psychophysical data, the model relates reasonably well to what is known
about the neurobiology of motion sensitive cells in the brain making it a realistic model of human visual motion sensing.
Some other highlights of the dissertation are as follows:
¿ I find that the intrinsic cortical noise in the brain which manifests itself as uncertainty
in motion estimation can play an important role in perception by significantly
improving detectability of subliminal motion cues at the expense of a very modest
drop in performance for a suprathreshold signal ala stochastic resonance.
¿ I also did experiments on observers under the influence of marijuana and found that
the THC in marijuana can cause an impairment of motion perception abilities ¿
observer performance decreases by as much as 15% and reaction time increases by
as much as 222±96 ms.
¿ I find that the observer performance is invariant to dot density in the display and argue
that this provides very powerful evidence against motion models based on matching
dots to nearest neighbors in successive frames ala (Ullman, 1979; Dawson, 1991)
etc.
¿ I find and prove that the rotary motion signal does not depend on the center of rotation
relative to which it is computed which explains the experimentally observed position
invariance of MST(d) cells found by (Graziano, Andersen, & Snowden, 1994).}
}

EndNote citation:

%0 Thesis
%A Jain, Siddharth
%T PSYCHOPHYSICS & COMPUTATIONAL MODELING OF VISUAL MOTION PERCEPTION
%I EECS Department, University of California, Berkeley
%D 2007
%8 August 7
%@ UCB/EECS-2007-97
%U http://www.eecs.berkeley.edu/Pubs/TechRpts/2007/EECS-2007-97.html
%F Jain:EECS-2007-97