Water objects extraction from polarimetric SAR imagery based on blind source separation and morphological reconstruction

The SOMMR nonlinear blind source separation (BSS) method is proposed for speckle noise suppression and water objects extracting from synthetic aperture radar (SAR) imagery based on self-organizing maps (SOM) neural networks and morphological reconstruction (MR). The multiplicative speckle noise and image data are separated from multipolarimetric imagery by means of SOM neural networks. Morphological reconstruction is employed to remove the residual noise. The experimental results using ENVISAT ASAR polarimetric imagery show that the proposed method can extract water objects accurately, and the speckle noise index is better than ICA and SOM method.