Date
29.03.2006

Time
13:15

Auditoriet, Nordlysobservatoriet

"Non-Stationary Analysis and Localisation in 2D" by Dr. Sofia C. Olhede

Dr. Sofia C. Olhede
Department of Mathematics, Imperial College London presents a guest lecture entitled
"Non-Stationary Analysis and Localisation in 2D"
on Wednesday 29 March, 2006, at 13:15 - 14:00 at Auditoriet, Nordlysobservatoriet

Everyone are cordially welcome!

ABSTRACT:

Image analysis has in the last decade experienced a revolution through
the development of new computationally tractable non-stationary analysis
methods. At the heart of these developments is the construction of
suitable local representations, or decompositions. The chosen
decompositions also form the basis for further analysis and/or estimation
methods. In particular, the compression of a particular representation
ensures that most decomposition coefficients are of negligible magnitude,
thus simplifying the analysis considerably.

A major issue is the choice of a suitable decomposition. To design
"optimal", or well-localised functions, a localisation operator is
constructed, and the eigenfunctions of this operator give a full set of
"optimal" orthogonal analysis functions. Using several functions has
several advantages, e.g., the estimation bias can be controlled, whilst
the variance is reduced.

The problem of developing optimal functions in two dimensions requires
defining a genuinely two dimensional localisation operator. We discuss the
choice of operator and also discuss using its eigenfunctions to estimate
local image characteristics. Finally, we demonstrate our method on data.

Kontaktperson
Professor Alfred Hanssen IF