Colloquium Computer Science: Dr. Oscar Dalmau Cedeno

Colloquium Computer Science

Title: Probabilistic segmentation and some applications

Abstract:

Image segmentation is a very important image processing task. This task can be divided in “hard” and “soft” segmentation. In this presentation we are interested in modeling the “soft” segmentation problem. We reformulate the probabilistic segmentation methods by using a variational approach and present a unifying framework for probabilistic segmentation. Our formulation is based on modeling the joint distribution of a random vector field (the observation) and a Hidden Markov Random vector field as a Gibbs distribution. Then, the problem of obtaining an estimator of the hidden Markov Random vector field is reduced to minimizing an energy function, and the probabilistic segmentation methods are obtained as particular cases when both random vector fields, the observation and the hidden Markov vector field, are restricted to be vector measure fields. The unifying framework allows us to easily obtain new probabilistic segmentation methods. For this purpose, we suggest to use metric-divergences between discrete distributions. Finally, we present applications os probabilistic segmentation to image editing and crops classification in satellite images.

Colloquium coordinators are Prof.dr. M. Aiello (e-mail : M.Aiello rug.nl ) and Prof.dr. M. Biehl (e-mail: M.Biehl rug.nl )

http://www.rug.nl/research/jbi/news/colloquia/computerscience