As part of a wider European study under the project MOUMIR
Models for Unified Multimedia Information Retrieval
http://www.mee.tcd.ie/~ack/moumir/
a Post Graduate Research studentship is available in Department of
Statistics, Trinity College Dublin http://www.tcd.ie/statistics/research/
_____________________________________________________________________________
We are all used searching databases, whether it is looking for a book by
subject or author in the library or looking for information on the
internet. All
these are text based searches and the information required is all in text.
There
are well established algorithms for performing such searches very quickly.
An alternative situation can exist when one has a database of images.
Newspapers and other TV and film broadcasting companies, and several art
galleries, now have such databases. Here, one might want to search for images
with certain properties (i.e. "give me an image with a beach at sunset").
This is a considerably more difficult task, requiring that (a) we are able
to define
the idea of a beach and a sunset to the computer and (b) that the computer is
able to look at images and assess which have the properties of a beach and a
sunset.
One approach to problem (b) above is through probability models for what
images look like. If one can define a sunset mathematically, the idea is
then to
analyse each image in the database and give a probability that it contains a
sunset. To do this requires many powerful tools in probability, including
Markov
chains, spatial probability models and stochastic optimisation techniques.
The project will require a student with good mathematical ability. A knowledge
of probability is useful but not essential. Much of the project will be
concerned
with writing code to implement these probabilistic approaches, and so a good
programming skills (in C/C++ preferably) are also important.
This is one of a number of projects within the wider study of the Highly
Structured Stochastic Systems research group in the Department of Statistics.
It will form part of a wider European Research Network with colleagues
in France, Greece, Isreal, Portugal and the U.K. Note that because of the
nature of this European project, applications can only be entertained from
citizens of the EU or accession states, and applicants may not be citizens
of the Republic of Ireland.
Further enquiries to [log in to unmask]
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
John Haslett E-mail [log in to unmask]
Professor Phone +353 1 6081114 (direct)
Department of Statistics +353 1 6081767 (sec)
Trinity College Fax +353 1 6615046
Dublin 2,Ireland
WWW: http://www.tcd.ie/statistics/people/johnhaslett.html
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|