1.1 - Optical flow
In
a sequence of images, object motion can be modeled by a continuous variation
of image intensity as a function of position (x,y) and time (t). Optical
flow computation refers to a technique by which motion, in each image
point, can be determined with respect to magnitude and direction. You
will study this technique in some detail and also try it on synthetic
as well as ultrasonic moving heart images. Optionally, you may use optical
flow for studying cell displacements in consecutive images acquired
by time-lapse microscopy. This problem is part of a current project
on stem cell tracking.
Contact: Peter Holdfeldt
1.2 -
Boundary Detection by Dynamic Programming
Dynamic programming (DP) is
an optimization technique for solving problems where not all variables
in the evaluation function are interrelated simultaneously. It has been
successfully applied to boundary detection in, for example, ultrasonic
imagery. Previously, we have successfully applied the DP algorithm to
horisonal boundaries. In this project, your task will be to
investigate modifications of the DP algorithm which can be applied to:
a) more
or less straight but non-horisontal boundaries.
b) boundaries
that are curved.
Contact: Peter Holdfeldt
1.4 - Characterizing arteriosclerosis by texture analysis
The Wallenberg Laboratory for Cardiovascular Research is interested
in the characterization of arteriosclerotic plaque in the human cartotid
arteries. Several files containing ultrasonic images presenting arteries
with plaque will be available. Your task is to develop a texture analysis
algorithm capable of discriminationg between different types of plaque.
Because many approaches can be applied, this project can be chosen by
more than one group.
Contact: Peter Holdfeldt
1.6 Cell segmentation
Segmentation of cells can be done using many different methods, e.g.
thresholding,
region growing or watersheds. The task is to implement, test, and assess
one or two of these methods. Specifically, you shoud adress the problem
of segmenting cells in clusters?
Contact: Peter Holdfeldt
