Abstract:

This paper presents a novel speckle reduction algorithm based on sparse representation and structure characters of the polarimetric synthetic aperture radar （PolSAR） image. Firstly, the image is classified according to the polarimetric properties to form the classification map. Secondly, the orthogonal matching pursuit （OMP） algorithm is applied to implement the sparse decomposition on PolSAR images, and the over-complete dictionary is updated by the K-singular value decomposition （K-SVD) algorithm to get the trained dictionary and sparse coefficients, with which the image without noise is reconstructed. Finally, the point and line targets are enhanced by the classification map in the reconstructed image. Experimental results with the data of the Hayward area from the AIRSAR system show that the proposed method is effective both on speckle reduction and scattering characteristics preservation.