A Temporal Estimation of Entropy and Its Comparison With Spatial Estimations on PolSAR Images

Abstract

Most of the applications of SAR polarimetry such as classification are based on estimation of the polarimetric covariance matrix. This estimation is generally done through a boxcar spatial filtering. This estimation process can induce mixture if different scatterers are present in neighboring pixels. Since the polarimetric entropy H is a measure of variability, this mixture can result in a very uniform entropy map. A nonlocal algorithm can be used to improve the estimation of the covariance matrices. The entropy maps are smoothed and contrast is better preserved. We propose a third estimation of H by using a temporal stack. Pixels are averaged on the time axis instead of on a spatial basis. On the datasets we studied, the temporal estimation increases the contrast of H maps. This contrast allows us to better discriminate targets. Temporal entropy is very influenced by the degree of coherence. Nevertheless, H <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">temporal</sub> provides additional information, combining information about the polarimetric stability of scattering mechanisms over time.