In this work several different approaches for the segmentation of Meteosat images are compared. The satellite images are segmented in the classes land (L), sea (S), fog (F), low clouds (CL), middle clouds (CM), high clouds (CH) and clouds with vertical growth (CV). All the approaches are based on neural networks and the differencesamong them are mainly at the pre-processing stage, in the selectionof the sutable set of features to carry out an optimal classification.