Abstract. The purpose of this paper is to present a system for
the analysis and classification of Synthetic Aperture Radar (SAR) images.
This system, unlike most of its competitors, allows a careful modeling
of the statistical properties of the data beyond the usual Gaussian hypothesis.
The modeling tools include basic descriptive measures and the choice of
suited distributions, through goodness-of-fit tests, to model the data.
The classification tools offer the choice between pointwise and contextual
(Markovian) techniques, and the quantitative assessment of the quality
of the results. The system is goal-driven, and its interfaces are solely
based on pull-down menus; the user is prompted with the correct sequence
of operations, whenever an invalid option is invoked. An example of the
use of this system for the classification of a SAR image is presented.
Keywords: classification, image, interface, radar, SAR, speckle,
system
[ full version | contact authors ]