STATISTICAL ANALYSIS OVER ARGENTINA AND ITS ATLANTIC 
PLATFORM TO BE APPLIED TO CLASSIFICATION TOOLS 
Laura Adriana Frulla 
J. Jacobo-Berlles 
Marta Mejail 
CAERCEM 
Julian Alvarez, 1218 
Buenos Aires, 1414, Argentina 
ISPRS Commission VII / Working Group 1 
ABSTRACT 
It is well-known that the statistical properties of the images depend on the type of surface that is being 
observed. But there is also an influence of the number of looks in combination with the surface 
characteristics. For example for one-look images over homogeneous areas, the density function of the speckle 
field is described by a Rayleigh or exponential distribution. But in general, due to the variations of the 
features of the earth surface, there is a wide type of theoretical distributions the hystogram can be adjusted to. 
It is very important to apply classification techniques to SAR images, but in general the classification models 
are developed for hystograms that can be adjusted to a multivariate normal distribution. In the case of SAR 
images the classification is based on bands which are not normally distributed. Thus it is necessary to derive 
other ways of classification. 
There is a lot of previous studies on the application of non gaussian noise model such as K and Gamma 
distributions. On the basis of these studies, the purpose of this paper is to use one look SAR imagery to study 
the statistical properties of different regions of Argentina problem of automatic classification of each type of 
region. Morever it is possible to analyse the error in using a classification algorithm for gaussian distributions 
when a one-look ERS-1/SAR image is being used. 
007