Speckle correction filter based on spatial polarimetric coherence for full polarimetric synthetic aperture radar image

Abstract

Polarimetric synthetic aperture radar (PolSAR) image exhibits a high degree of spatial information. However, the PolSAR image is subjected to speckle noise, which degrades the information quality. A speckle correction filter is proposed in this paper based on spatial polarimetric coherence. Evidence of polarimetric coherence between spatial neighboring pixels is presented. It shows that polarimetric coherence is highly correlated with neighboring pixels. Based on these characteristics, a speckle correction filter is proposed by calculating the polarimetric coherency level on a selected neighboring pixel. A correction is applied to the pixel which has a low polarimetric coherence level. A generalized three-step adaptive filter is proposed, including neighbor selection, coherency criterion, and incoherent correction. The influence of the filter on scattering mechanisms is compared by using eigenvalue-based decomposition. Qualitatively, the result shows the proposed correction filter removes the outliers in the low entropy region. Quantitively the filter can suppress the speckle noise and its variance up to 19.08%.