A novel ICA-based method for image enhancement of polarimetric SAR
Xuwu Su, Hongshi Sang, Guangyou Yang, Shen Xu-bang · 2011
A new image enhancement method applied to polarimetric SAR image is proposed in this paper. Instead of widely-used intensity image, this method processes real part and imaginary part of complex SAR image. Independent Component Analysis (ICA) algorithm is used to extract enhanced real and imaginary images, for multiplicative speckle noise in intensity SAR image is additive in complex domain according to speckle forming principle. Another reason is that speckle noise in real or imaginary image is Gaussian distributed while heterogeneous area has non-Gaussian distribution. The correlations of all real and imaginary images are studied. The highly correlated polarization channels are HV and VH, HH and W, which implies where independent components should extract from. After ICA processing, the extracted real and imaginary images show some improvement, and the reconstructed intensity images indicate distinct enhancement in image contrast. The deficiency of this method is partly lose of polarimetric information. The sample POLSAR image is from Radarsat-2.