Open AccessStokes based sigma filter for despeckling of compact PolSAR data
Author(s) -
Sharma Rakesh,
Panigrahi Rajib Kumar
Publication year - 2018
Publication title -
iet radar, sonar and navigation
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.489
H-Index - 82
eISSN - 1751-8792
pISSN - 1751-8784
DOI - 10.1049/iet-rsn.2017.0313
Subject(s) - speckle pattern , filter (signal processing) , stokes parameters , polarimetry , computer science , wishart distribution , synthetic aperture radar , algorithm , speckle noise , artificial intelligence , pattern recognition (psychology) , scattering , physics , computer vision , optics , multivariate statistics , machine learning
In this study, a speckle filtering technique is proposed for coherent compact polarimetric synthetic aperture radar (PolSAR) data based on the statistics of Stokes parameters. The data processing and analysis of coherent compact PolSAR configurations usually start with Stokes vector representation of the data. In this study, the Stokes parameter statistics is discussed and validated, and a modified form of the state‐of‐the‐art Lee sigma filter for compact PolSAR data is presented. Also, a novel algorithm for evaluation of sigma ranges as a function of the degree of polarisation and mean intensities is proposed. An approach based on the statistics of Stokes parameter g 1 for the distinction of homogeneous and heterogeneous areas is also presented in this study. The proposed technique is effective in speckle reduction, preservation of fine details, strong targets and scattering information. The m − χ decomposition with the proposed filter as a pre‐processing step improves the land cover classification accuracy. The computational efficiency of the proposed filter is comparable to Lee sigma filter and hence the proposed filter is equally suitable for voluminous high resolution datasets. The results obtained from the developed speckle filter are validated for hybrid‐PolSAR RISAT‐1 data, hybrid‐PolSAR data synthesised from RADARSAT‐2 data and simulated PolSAR data by Monte Carlo simulation.