\magnification=1200
\baselineskip=20pt
\nopagenumbers
\font\big=cmr12 scaled \magstep2

\centerline{\bf STANFORD UNIVERSITY}
\centerline{\bf DEPARTMENT OF STATISTICS}
\centerline{\big DEPARTMENTAL SEMINAR}
\bigskip

\baselineskip=12pt
\centerline{4:15 p.m., Tuesday, September 30, 2003}
\centerline{Sequoia Hall Room 200}
\centerline{(Cookies at 3:45 in 1st Floor Lounge)}

\bigskip
\baselineskip=15pt
\centerline{\sl Rasmus Larsen}
\centerline{\sl Informatics and Mathematical Modelling}
\centerline{\sl Technical University of Denmark}
\bigskip
\centerline{\bf Biomedical Image Analysis - Analysis of Biological Appearance}
\bigskip

Abstract:

I will present the active appearance model (AAM) proposed by
Cootes \& Taylor for modelling the appearance of biological
objects in images. This model is based on a truncated series of
principal components and it is trained on images of biological
objects (manually) annotated by sets of landmarks (corresponding
points). I will illustrate the use of AAM for modelling and
segmentation of images by examples including the human face,
cardiac structures and brain structures. I will discuss 2 problems
in relation to the application of AAM, 1) manual annotation of
landmarks is tedious and error prone in 2D and insurmountable at
realistic image resolutions in higher dimensions. Alternatively,
training input may consist of curves and surfaces for which point
correspondences can be established; 2) segmentation using AAM is
based on relating the difference image between model and image to
a change of parameters. For moderate resolution images in
dimensions higher than 2 this becomes computationally infeasible.
We therefore investigate the use of truncated wavelet and
wedgelets bases to represent the images.

\bye