Structure preserving bilateral filtering for PolSAR data

This paper proposes a structure preserving bilateral filtering (SPBF) for Polarimetric SAR (PolSAR) image. By adopting the edge structural characteristics and surface scattering features, the performance of filter is improved and the loss of structure is reduced. To begin with, the edge direction is determined by edge templates and an adaptive direction window is selected in span image. Then each scattering mechanism of the pixel is attained by Freeman-Durden decomposition. The surface scattering map is obtained by statistic distribution characteristics of the polarimetric data. Finally, the filtering mask, combining the cluster map with adapted direction window, is introduced for an improved bilateral filtering. Experiments illustrate the effectiveness of the proposed method. The speckle is efficiently reduced and the image edge, strong point target and polarimetric scattering characteristics can be further preserved.

1. Introduction

PolSAR is a multi-dimensional imaging system with all-day, all-weather and high-resolution. PolSAR backscattering returns can be represented as HH, VV and HV polarizations for distributed targets. Due to the correlations of polarimetric channels, speckle effect affects not only three intensities, but also three phases [1]. To reliably interpret PolSAR, the speckle must be reduced.