A new method for full-polarization SAR images classification integrated with speckle reduction is proposed.SAR image is firstly decomposed by Pauli-decomposition approach and three combined polarization channels which are denoted by three different colors are obtained.Then the speckle of each color channel is reduced by using ICA sparse-coding shrinkage algorithm (ICA-SCS).After speckle reduction

the three channels are constituted to rebuild a classified image with less speckle noise.This method well keeps the relative phase information among the polarization channels.Furthermore

the channels after speckle reduction are directly used for classification without any other combination.The classification results of real SAR images indicate that the classification accuracy is improved obviously.