Multiplicative multiresolution decomposition for 2D signals: application to speckle reduction in SAR images

The new concept of multiplicative multiresolution decomposition (MMD) with perfect reconstruction is presented. This kind of analysis/synthesis representation is suitable for images corrupted by multiplicative noise, e.g. synthetic aperture radar (SAR) images. The design and implementation of a nonlinear multiplicative filter bank are addressed. An application to speckle reduction in SAR images is also described. The performance of the MMD speckle reduction method is studied and compared with those of the most often used adaptive filters.