Feature preserving reliable estimate of the polarimetrie coherency matrix using IDAN-LLMMSE technique

摘要

The data of Synthetic Aperture Radar (SAR) is affected by speckle noise due to the coherent integration of back scattered signals from different targets. The speckle filter of any kind has to suppress the speckle noise while preserving the polarimetric and the spatial information. Speckle filtering of Polarimetric Synthetic Aperture Radar (POLSAR) data remain a challenging task because of the difficulty in reducing a scattered-dependent noise. The filtering technique of Intensity Driven Adaptive Neighbourhood (IDAN) for each pixel of the data will set an adaptive neighbourhood that is computed by a region growing technique driven exclusively by the intensity images. The acquired three intensity images of the POLSAR data are fused in the process of region growing to ensure the stationarity hypothesis of the derived statistical population. Then, pixels present within the adaptive neighbourhood are estimated or complex averaged. The general remarks for all the spatial filters are the resulting filtered images which have a patchy look and known effect of purely spatial filtering. This effect is also observed for the IDAN. The estimation is done by using Locally Linear Minimum Mean-Squared Error (LLMMSE) instead of complex averaging. The IDAN-LLMMSE yields the reliable estimation of the polarimetric coherency matrix when compared to IDAN.