A novel despeckling algorithm of polarimetric SAR image based on SNR and parameter-vector spectrum amendment

The subspace filtering of polarimetric SAR image can retain polarimetric signature very well, but for its limitations it is necessary to enhance its despeckling effect. So a novel despeckling algorithm of polarimetric SAR image based on SNR and parameter-vector spectrum amendment is proposed. First, the sufficient and necessary condition of signal-noise separation based on subspace filtering is analyzed to conclude that signal is not completely separated from noise, so a new despeckling tactic based on SNR is produced, which means preserving components with high SNR and discarding components with low SNR. Then, for the defect of reconstructing parameter vector in the subspace filtering, a method of parameter-vector spectrum amendment is derived, which can enhance despeckling effect and keep the filtered polarimetric data close to the unfiltered. At last, the experiment shows that the novel algorithm definitely has better despeckling effect than the subspace filtering, and retains edge-point feature and polarimetric signature as well.