Abstract：In this paper an automatic multiresolution classification method is proposed to classify polarimetric synthetic aperture radar (SAR) image.At first a textured SAR image characterizing the terrain reflection is obtained by using a multi-look polarimetric whitening filter (MPWF) to reduce the speckle in the polarimetric SAR image. Then the wavelet transform (WT) is utilized to extract texture information in different resolutions, and in the lowest resolution level the Akaik information criterion (AIC) is used for estimating the optimal number
of texture classes in the image. Next the Markov random field (MRF) model is employed to characterize the spatial constraints between pixels in each resolution level, and a maximum like-
lihood (ML) approach and an iterated conditional mode (ICM) approach are used for the model parameters estimation and maximum a posteriori (MAP) classification, respectively. Finally the paper presents the experimental results with the NASA/JPL L-band airborne polarimetric
SAR data and verifies the effectiveness and advantage of the classification method proposed.
2000, Vol. 22 
