My presentation will focus on the comparison between spatially estimated polarimetric entropy and temporally estimated polarimetric entropy. Indeed, long term temporal series become accessible for SAR images in full-polarimetric mode, thanks to the new lauches of satellites and to the wider distribution of SAR images by spatial agencies. It allows to monitor the evolution of scattering mechanism in time and to use different tools to characterize the scene.
Polarimetric entropy is a classical tool for PolSAR images study and classification. It is often used as an indicator of the variability of scattering mechanisms for given homogeneous areas. In order to estimate entropy, the covariance matrix has to be estimated. This is mostly done using the sample covariance matrix estimator on adjacent pixels. Thus, the entropy becomes an indicator of the variability of scattering mechanisms present in the sample set used for its estimation.
Using a temporal series, new estimation methods of entropy can be employed. For example, entropy can be estimated using the temporal samples of the same pixel. This temporal entropy gives us information about the temporal stability of the scattering mechanisms. New classes of scatterers seem to emerge in natural areas. In urban areas, the contrast of entropy between streets, parks and buidlings is less dependent on the street orientation with temporal entropy than with the classical spatial entropy.
These results on entropy estimation open new perspectives for an Entropy/alpha classification scheme combining spatial and temporal entropy.