You are welcome to attend the following. Please also publicise among
your colleagues. Attendance is free.
MEETING ON OPTIMISATION METHODS
Joint meeting of the Royal Statistical Society Statistical Image
Analysis and
Processing Study Group (SIAP) and
the Avon Local Group of the Royal Statistical Society
Friday 19 October, 2.15pm- 4.45pm (tea 3.15pm)
University of Bristol, Department of Mathematics Seminar Room SM3
(local Website http://www.stats.bris.ac.uk/Seminars/)
Riccardo March (Istituto per le Applicazioni del Calcolo, CNR, Rome)
Variational Problems in Image Analysis and the
$\Gamma$-convergence Method
The Mumford-Shah variational model for image segmentation is
introduced, and the application of the variational principle
to other vision problems, such as recovery of depth from stereo images,
is discussed. The resolution of the corresponding variational problems
by means of standard numerical methods of calculus of variations is
difficult because the solution is discontinuous along unknown
curves (which typically correspond to object boundaries).
The $\Gamma$-convergence method is then introduced, which allows
the approximation of such variational problems by means of simpler
problems which can be solved numerically using standard finite
difference/finite element methods. The Ambrosio and Tortorelli's
approximation is presented together with a simple finite difference
algorithm. Numerical examples of segmentation of real images and stereo
reconstruction from synthetic images are shown for illustration.
Christophe Andrieu (University of Bristol)
Improved Particle Filter Techniques with an Application to
the On-line Model Selection and Estimation of TVAR
In this work we propose new efficient particle filter
techniques which prove to be key to the success of
on-line model selection and estimation for TVAR processes. Results on
real data sets will be presented. Some thoughts about how
these particle filter techniques may be applied to image analysis
will be given. This is joint work with Emmanuel Davy and Arnaud Doucet.
|