Study of information content of SAR images

Abstract

Recent and future space missions for Earth Observation grant a growing avaibility of high resolution SAR images. The analysis of this data tipology critically requires exploiting the spatial-contextual information in the related images. In this paper we propose a contextual measure of the amount of information in a SAR image based on Gauss Markov Random Fields, which takes into account both the radiometric resolution and the spatial one. This is related to an analitical expression of the channel capacity associated to the SAR image, i.e. a channel affected by a Gamma distributed multiplicative noise.