ICA based PolSAR image denoising
Raul Malutan, Romulus Terebeş, Mihaela Cislariu, Christian Germain, Ioana Ilea · 2016
A speckle denoising in polarimetric SAR images is proposed considering as filtering method the sparse code shrinkage algorithm based on ICA. Even thou the ICA method was widely used in audio signal processing, in combination with shrinkage function give good results for multiplicative noise reduction. The Extended Generalized Lambda Distribution (EGLD) algorithm was set to estimate the transform matrix before applying the sparse coding. The proposed method was tested on PolSAR datasets and compared with well-known speckle noise removal algorithms. The evaluation indexes gave good results for the ICA based method.