--------------------------------------------------------------------------


		     B O S T O N   U N I V E R S I T Y

			Computer Science Department

			    C O L L O Q U I U M

	     Tracking Figures in Video Using Kinematic Models

				 Jim Rehg
			  Cambridge Research Lab
			   Compaq Computer Corp


			  Wednesday, September 30
				  3:00pm

			 (Coffee served at 2:45pm)

			  Seminar Room / MCS 135


Noninvasive techniques  for measuring human motion are  important in a wide
variety of  applications including computer  graphics, user-interfaces, and
surveillance.  Computer vision   techniques  are attractive  because   they
utilize   inexpensive cameras and  can  be applied   to existing multimedia
content such  as video footage.  Visual  analysis of figure and hand motion
is  difficult because of   its  nonrigidity, large   numbers of degrees  of
freedom, complex  self-occlusions   and  variations  in  appearance.  These
difficulties can be   overcome   through  the  use of  kinematic    models.
Kinematic  models provide a compact representation  of motion and constrain
image interpretation.

I  will describe  a  framework for  visual  tracking of articulated objects
using 3-D  and 2-D  kinematic   constraints in conjunction  with  a layered
template  appearance model.   This  framework does not  depend  on an exact
shape model of the tracked object and  encompasses both single and multiple
camera configurations.   I will show three  applications of this framework:
real-time hand  tracking (the DigitEyes system),  figure tracking in movies
(the Virtual  Fred Astaire project), and figure   detection.  This is joint
work with Tat-Jen Cham, Daniel Morris, and Henry Rowley.

Host: Prof. Stan Sclaroff (sclaroff@bu.edu)

---------------------------------------------------------------------------
For colloquium info and directions, see http://www.cs.bu.edu/colloquium
---------------------------------------------------------------------------