Minimal Stochastic Complexity Image partitioning with nonparametric statistical model
Guillaume Delyon, P. Réfrégier, Frédéric Galland · AIP conference proceedings · 2006
We propose to analyse a general statistical image partitioning into homogeneous regions. This method is based on a polygonal grid which can have an arbitrary topology and whose number of region and regularity of its boundaries are obtained by minimizing of the stochastic complexity of the image. It thus leads to optimize a criterion that can be expressed as the sum of two terms: a likelihood term and a regularization term. This criterion, whose two terms are elements of a global entropy and are expressed in the same unit, does not contain any parameter to be tuned by the user. We analyse the performance of this approach in comparison with a more standard formulation for which there is a weighting parameter between the likelihood and the regularization terms analogous to the one used in snake‐based techniques.