An Improved PolSAR Image Speckle Reduction Algorithm Based on Structural Judgment and Hybrid Four-Component Polarimetric Decomposition

Abstract

Aiming at the structural feature and polarimetric property preservation in speckle reduction of polarimetric synthetic aperture radar images, an improved speckle reduction algorithm based on structural judgment and hybrid four-component polarimetric decomposition (HPD) is developed. To protect the structural feature, a new structural judgment based on edge masks and pixels' sort is used for dividing pixels into three structural categories: bright, dark, and common targets. To protect the polarimetric property, a new scattering classification with HPD is employed to divide further the dark and common targets into four scattering classes. HPD inherits the physical significance of Pauli and Yamaguchi decomposition with its advantage of less time consumption. After that, without bright targets, a dark or a common pixel centered in a sliding window is filtered. To achieve better filtering results, pixels in its same and three neighboring scattering clusters from the same structural category are selected to filter. The experimental results of the UAVSAR and RADARSAT-2 data illuminate the validity of our proposed method.