Change Detection Analysis using Information Theoretic Measures on SAR Images

We discuss the use of two statistics for change detection in the context of Synthetic Aperture Radar (SAR) imagery. We show their application to a bi-temporal pair of HH and VV channel intensity images from RADARSAT-2 of an agricultural scene in Winnipeg, Manitoba, Canada. The images were acquired on 7th and 31st July 2012. One of the statistics is based on a stochastic distance viz., Hellinger distance, while the other is based on the Shannon entropy which provides a measure of randomness. We have assumed the Gamma model for the HH and VV channel intensity data with mean and number of looks as the two free parameters. Test statistics are often used to design hypothesis tests after theoretically deriving their asymptotic distributions. Such hypothesis tests, to have practical utility, require a judicious choice of sample size, and a level of significance for thresholding respectively. Instead of relying on the validity of the asymptotic distribution of test statistics (i.e. an implicit assumption for using p-values and levels of significance), in this work we use the test statistics as direct quantifiers of change. We apply a simple k-means clustering with k = 2 to these quantifiers in order to segregate change and no-change regions. With these, we show that the both information theoretic measures provide substantive evidence for change detection. The corresponding change maps are studied together to understand the complementary nature of the selected statistics. It is inferred that these two statistics may be used in tandem for better change detection analysis in SAR imagery.