The considered classes consist in all those obtained with the CVC on the initial data ( field classes ) plus one extra class corresponding to urban areas and defined on the textural parameter class with the highest grey levels. We do not consider the other classes detected on the textural parameter because they are not reliable. This knowledge is the own supervised part in the whole algorithm.
Parameters associated with each class ( mean and variance ) are computed on the initial data and the textural parameter using the classifications obtained from the CVC.
The data attatchment in the relaxation process correspond to the first order potentials of the Markov Random Field. They are written as follows :
SYNOPSIS OF THE DEMO 