Aarhus University Seal / Aarhus Universitets segl

Identifying fibres in spatial point patterns

Bryony Hill
(University of Warwick)
Onsdag, 12 maj, 2010, at 14:15-15:00, in Koll. G3 (1532-218)
In this talk I'll give an overview of a Bayesian analysis of the distribution of systems of fibres arising from noisy point processes (or point patterns). A fibre (or curve) in this context is a one-dimensional object embedded in the same space as the point process. Examples of point patterns with a fibrous structure exist in a numerous places in nature: earthquakes along fault lines, galaxies clustering around filaments, sweat pores lying along fingerprint ridgelines, etc.

Our approach to the problem involves identifying fibres as streamlines of a gradient field, and using a Birth-Death MCMC algorithm to sample from the posterior fibre distribution. The gradient field is calculated as part of an Empirical Bayes process, using tensor methods commonly seen in Diffusion Tensor Imaging.

Organiseret af: CSGB
Kontaktperson: Eva Bjørn Vedel Jensen