a method for unsupervised terrain classification of POLSAR imagery is proposed.At first

the polarimetric entropy is utilized to divide terrain scattering into three cases

i.e.high entropy

medium entropy and low entropy.Then surface scattering

double scattering and volume scattering similarities are computed for initially classifying POLSAR image into ten classes.Finally

the initial classification map defined training sets for reclassification based on a new defined parameter to measure the deference degree between two targets.As surface scattering

double scattering and volume scattering are the inherent characteristics of terrain physical scattering

the scattering classified results are more accord with real terrain scattering;the automatic determination of scattering type with scattering similarity overcomes the deficiency of the present scattering classification;the simple computation of the deference degree speeds up the above reclassification.The experiment results with real POLSAR image demonstrate the validity of the proposed method.