Bayesian methods in content-based image retrieval

Speaker: Dr. Simon P. Wilson (Statistics Department Trinity College Dublin)

Time: 12:00AM
Date: Fri 13th October 2006

Location: Mathematical Sciences Seminar Room

Content-based image retrieval refers to the indexing and retrieval of images from a database using their semantic content. In this talk some statistical solutions are presented that use Bayesian methods. Three issues will be addressed: how to produce a learning algorithm (via Bayesian inference), which images to select from the database to show to the user (via decision theory) and how to use data from previous similar searches to improve performance (again via Bayesian inference). A demo will be shown based on a database of paintings.

(This talk is part of the Statistics and Actuarial Science series.)