Bias characterization in H/A/α polarimetric SAR decomposition and its effect for the classification

摘要

Entropy, anisotropy and alpha (H/A/α) parameters of the target decomposition in synthetic aperture radar polarimetry (PolSAR) are valuable tools for the assessment of the physical parameter retrieval. However, the speckle emerging in SAR images decreases the accuracy of image classification and segmentation. Consequently, it should be filtered correctly. In fact, insufficient averaging generates biased estimates of H/α/A parameters and overaveraging degrades the resolution cell. In this paper, we investigated bias estimation deeply. We enhanced bias compensation of the entropy. The implication of bias for PolSAR data classification has not been studied yet. We demonstrated that bias have an important consequences on PolSAR data classification. Real and simulated data are used.