[4eyes] Four Eyes Lab meeting tomorrow, 5:00pm

[Matthew Turk]

We have a lab meeting tomorrow at 5:00pm, in the MAT conference room (2nd
floor Elings Hall). Sharath will talk about the state-of-the-art in shape
representation and matching. We will also discuss plans for an
end-of-the-year party.

See you there,

	Matthew



Shape Matching

A central problem in Computer Vision and Pattern Recognition is to
determine the extent to which one shape differs from other. This has many
applications in fields of medicine and in entertainment industry. It is
one of the most active and wide area of research, with literature dating
as early as 1917.

The state of the art in shape matching has moved beyond the traditional
pattern recognition approaches such as correlation and template matching.
This is due to the complexities introduced due to occlusions,
deformations, and in general, the need to match 'similar' objects. Today,
most of the approaches can be differentiated by three factors namely
(a) Feature extraction and representation of shape
(b) Solving correspondence problem by feature matches/Modeling the shape
(c) Determining the closeness of match/Recognizing the shape

I will briefly introduce the traditional methods of matching shapes and
then talk about a few modern approaches like Shape Contexts, Graph
Matching methods and methods dealing with deformation/distortion. I will
also cover recent approaches which solve the problem by learning the shape
models.