Parametric autofocus of SAR imaging - inherent accuracy limitations and realization

In synthetic aperture radar (SAR) imaging, low scene contrast may degrade the performance of most of the existing autofocus methods. In this paper, by dividing a slow-time signal into three isolated components, namely target, clutter, and noise, in SAR imaging, a novel parametric statistical model is proposed during the coherent processing interval. Based on the model, Cramer-Rao bounds (CRBs) of the estimation of unknown parameters are derived. It is shown that the CRBs of the target parameter estimation strongly depend on the background, i.e., clutter and noise, and the CRBs of the background parameter estimation may be obtained regardless of the target component. Motivated from this result and using the estimated background parameters, a novel effective parametric autofocus method is developed, which is applicable to any scene contrast. In addition, a preprojection is also introduced to simplify the subsequent parameter estimation. Finally, the proposed model and the novel method are illustrated by some real SAR data.