Edge Preserving SAR Image Despeckling

Synthetic Aperture Radar (SAR) images are inherently affected by multiplicative speckle noise, which is due to the coherent nature of scattering phenomena. Though a plethora of techniques are available to deal with speckle noise and SAR images, preserving edge details while despeckling still remains an issue. This paper models speckle as additive Gaussian in log space and proposes a wavelet based despeckling technique to preserve the sharp features and details of the SAR images effectively. Edge components are isolated in wavelet domain and a set of filtering scheme is applied on each of edge and non-edge components independently. Filtering scheme uses a polynomial based soft thresholding to remove the noise. Experimental results demonstrate that the proposed method performs better than existing Filters like Median, Lee, Kuan etc.