Sep 10, 2001
A Dissipative Approach to the Figure/Ground Problem
Dr. Mary Pugh, University of Toronto
In this talk I will present some work in progress in vision
research. We consider the problem of recognizing what parts of an
image are perceived as being in the foreground. We use a variant of
the Pao-Geiger-Rubin model, which uses an energy dissipation approach
to this problem. The model is surface-based, rather than
contour-based. Specifically, the edges in the image are not viewed as
isolated contours, but are viewed as bounding a surface. Each local
edge has a local hypothesis; for example, a north-south edge might
think "the region immediately to the left of me is part of the
figure". The model then uses energy dissipation methods to seek
assignments of local hypotheses that are mutually agreeable, yielding
a segmentation of the image that might be perceived. We test the model
on various images to address questions like: does the model "perceive"
smaller objects to be in the foreground (the way we do)? convex
objects to be in the foreground (the way we do)? how does it perform
on optical illusions that viewers report to have two different
segmentations?
This is joint work with Nava Rubin of the Center for Neural Science,
NYU. I thank Anita Disney (CNS, NYU), Davi Geiger (Courant, NYU), Bob
Shapley (CNS, NYU), and Dave McLaughlin (Courant, NYU) for useful
discussions.
Audio for this talk requires sound hardware, and RealPlayer or RealAudio
by RealNetworks.
Begin WebCam and audio for the whole talk: high bandwidth or medium bandwidth.
Or, begin audio only for the whole talk:
high bandwidth or low bandwidth.
(Or, right-click to download the whole audio file.)
To begin viewing slides, click on the first slide below.
(Or, view as pdf.)
Author entry (protected)