Date(s) : 04/04/2013 iCal
14 h 00 min - 15 h 00 min
A Markov point process for Fiber extraction\nBy Claire Coiffard\, IRMA\, Unistra.\n\n\nAbstract:\nThe aim of this work is to extract fibers in an image. We use a stochastic modelisation of the image : an observation is a realisation of a marked process. Using a bayesian approach\, we define our estimator by a MAP criterion. The model used to define the density of our process is the Gibbs process. The energy associated is divided into two parts : an internal field that favors specific spatial structures and the external field that quantifies the quality of the configuration with respect to the data. The problem of optimization is solved by a simulated annealing and MCMC methods.