Using Biquarternions Algebra and Joint Time Frequency Aanalysis Towards a New PolSAR Data Indexing Method

Abstract

Considering the multitude of nowadays high resolution remote sensing sensors, the volume of the acquired data has grown considerably. A particular type of the so-called "big data" is the polarimetric synthetic aperture radar (PolSAR) data. Processing this kind of data can be carried out by using either a parametric or non-parametric approach. This paper follows a new patch-oriented non-parametric approach. The BiQuaternion Fractional Fourier Transform (BiQFrFT) and the Method of Logarithmic Cumulants (MoLC) have been used as mathematical support. BiQFrFT was derived from the Fractional Fourier Transform (FrFT) and the biquaternions algebra. Representing PolSAR data samples in a rotated joint time-frequency plane via FrFT provides a simple statistical response that is easier to analyze. The biquaternions algebra allows the joint use of the polarimetric channels, enabling the polarimetric correlations between PolSAR data's samples. The logarithmic cumulants of the BiQFrFT coefficients were employed to build powerful, compact feature descriptors. The performances of our algorithm were tested on a spaceborne L-band PolSAR dataset. Finally, the better performances of our algorithm were shown against the ones of the well-known unsupervised Wishart H/alpha classification algorithm and some conclusions were drawn.