Approximation of distributions for SAR images: proposal, evaluation and practical consequences

Abstract

Data from a coherent illumination system like a Synthetic Aperture Radar (SAR) can be modelled by the multiplicative model. This model explains the characteristics of these data. The multiplicative model states that, under certain conditions, the constructive and destructive interferences of the incident and reflected signals give a return that varies, in a random manner, with the mean value of the backscatter corresponding to the illuminated target. The K distributions have been widely used to model data comming from a coherent illumination system, in particular for SAR images. These distributions allow us to model homogeneous as well as heterogeneous data. Given the need of finding a model for extremely heterogeneous data, a new class of distributions, the G distributions and a particular case the GA0 distributions have been recently proposed. These distributions model urban area very well, while K distributions do not fit this type of data. In this work, the feasibility of substituting the GA0 distribution for the KA distribution is studied and the practical usefulness of doing this will be shown. Once this fact is established, it will be possible to consider SAR images under the GA0 model only.