This paper describes the application of neural network techniques to t
he automatic segmentation of Synthetic Aperture Radar (SAR) images. Tw
o techniques have been studied : the Multilayer Perceptron (MLP) and t
he Learning Vector Quantization (LVQ). For an efficient classification
, the neural network is applied to a set of features extracted from th
e SAR image. Examples of suitable features are the radiometric informa
tion, the spatial frequency and so on. Two techniques are studied for
feature extraction : the Gabor filter and the Principal Component Anal
ysis (PCA). The paper reports the study in terms of correct classifica
tion rates related to four classification schemes obtained by a suitab
le combination of the feature extraction stage and neural network stag
e.