[Statlist] Seminar ueber Statistik

[Christina Kuenzli]

                       ETH and University of Zurich 
                                 Proff. 
A.D. Barbour - P. Buehlmann - F. Hampel - H.R. Kuensch - S. van de Geer

**************************************************************************     
           We are pleased to announce the following talk
**************************************************************************

Friday, February 3, 2005, 15.15, LEO C 15

       Complexity Penalized Segmentations: Efficient Algorithms and 
       Applications of Wedgelet-Type Image Approximation

       Felix Oliver Friedrich
       Institut für Computersysteme, ETHZ

We present a highly efficient method for the computation of
wedge-type approximation of digital images and an application with
respect to image  denoising.
We will address image segmentations with quadtree structures which have 
been first introduced by Donoho. A criterion for the choice of such a 
segmentation is given by minimization of a complexity penalised 
functional over the class of dyadic wedgelet partitions. Due to the 
structure of the functional, the minimization problem can be reduced to 
local comparison of scores obtained by moment computations over each 
partition segment. The performance of the method depends crucially on 
the fast computation of moments over wedge-like domains. We present a 
generic algorithm for fast moment computation over polygonal discrete 
domain.
The dramatic savings in computing time, which result from the new 
techniques, open new fields of applications for wedgelet approximations.
Some applications of the fast method with respect to multi-scale image 
denoising will be presented.
The algorithms have all been recently implemented in Oberon. We will 
illustrate our theoretical results with software demonstration.
     
________________________________________________________
Christina Kuenzli            <kuenzli at stat.math.ethz.ch>
Seminar fuer Statistik      
Leonhardstr. 27,  LEO D11      phone: +41 (0)44 632 3438         
ETH-Zentrum,                   fax  : +41 (0)44 632 1228 
CH-8092 Zurich, Switzerland        http://stat.ethz.ch/~