The Short-Time Beltrami Kernel for Despeckling Polarimetric SAR Data

Abstract:
In this paper, we adapt the Short-time Beltrami (STB) kernel used to smooth manifolds or images painted on manifolds to use the covariance matrices found in polarimetric synthetic aperture radar (PolSAR) data to reduce speckle noise. The proposed method functions similar to a region growing algorithm finding similar covariance matrices and weighting them accordingly. Preliminary results from the experiments show that the proposed method delivers a good compromise between the preservation of the spatial resolution and smoothing of homogeneous zones.