SNR and Noise Variance Estimation in Polarimetric SAR Data

Abstract

The problem of estimating the signal-to-noise ratio (SNR) of the cross-polarised channels and the noise variance in polarimetric synthetic aperture radar (SAR) data is dealt with. The Cramer-Rao Lower Bound (CRLB) is evaluated and a maximum likelihood (ML) estimator is derived, which jointly estimates the SNR of the cross-polarised channels and the noise variance. The performance of the joint estimator is assessed and a comparison with a coherence-based (CB) SNR estimator and an eigenvalue-based (EB) noise variance estimator is carried out. As far as the SNR estimation is concerned, both the ML and the CB estimator are biased, but the bias of the ML estimator is smaller than the bias of the CB estimator, while the accuracies are very similar. As far as the noise variance estimation is concerned, the ML estimator is unbiased and its variance is equal to the CRLB, while the EB estimator is biased. The difference in the biases is also shown using TerraSAR-X fully-polarimetric data, acquired during the Dual Receive Antenna (DRA) campaign.