Maximum Likelihood Estimation of Doppler Centroid in Synthetic Aperture Radar Images

This paper addresses the problem of estimating the DC (doppler centroid) for SAR (synthetic aperture radar) data, in presence of speckle and thermal noise. The main idea is to exploit a bandwidth much wider than the PRF (pulse repetition frequency), say 3-5 times, by exploiting strong point targets. Natural and isolated targets with close-to-ideal features are focused using a digital spotlight algorithm applied to stripmap acquisitions; then, accurate DC is estimated as the relative shift of the targets azimuth spectra using the ML (maximum likelihood). The Cramer-Rao bound of the estimate is shown to be lower than that of the conventional estimators based on traditional stripmap focusing of Q + 1 times for high SNR (signal-to-noise ratio) of targets, being Q the oversampling factor of the spotlight processing compared to the stripmap one. The method is applied to both simulated targets and a real dataset coming from the Cosmo SkyMed X-band constellation