Applications of ICA for Filtering of Fully Polarimetric SAR Imagery

A new class of filters for fully PolSAR image using ICA algorithm is proposed in this paper.This method,first transforms the four channel dB level intensity images by ICA,and gets four independent components,then sets noise components to the minimum,third utilizes the mixed matrix to calculate the four intensity image without noise,finally uses the original phase to work out the scattering matrix haven been filtered speckle.For verifying the validity of the method,the EMISAR data of Foulum is tested.Experiment results demonstrate that the method proposed can effectively filter speckle,and also has many advantages in preserving detail information.Even though polarization signatures of the scattering matrix after filtering have changed,differences and distinctive of objects are kept.Therefore,the proposed filter has large potential application in the fields of classification,edge and small target detection based on the statistical characteristics.