Predictive Quantization for Staggered Synthetic Aperture Radar

For upcoming spaceborne SAR mission the amount of data collected onboard is increasing, due to the employment of large bandwidths, multiple polarizations, and large swath widths, which lead to hard requirements in terms of onboard memory and downlink capacity. In this context, SAR raw data quantization represents an essential aspect, since it affects both the amount of data to be stored and transmitted to the ground and the quality of the resulting SAR products. In this paper, a data reduction approach based on predictive quantization is investigated in the context of Tandem-L, a DLR proposal for a highly innovative bistatic L-band radar satellite mission, aimed at monitoring the dynamic processes of the Earth. The proposed technique takes advantage of the time-variant autocorrelation properties of the non-uniform azimuth raw data stream in order to reduce the amount of data through a novel quantization method, named Predictive-Block Adaptive Quantization. Different prediction orders are investigated by considering the trade-off between achievable performance and complexity. Simulations for different target scenarios show that a data reduction of about 17.5% can be achieved with the proposed technique with a modest increase of the system complexity. Moreover, having a priori information on the gap positions in staggered SAR, a technique for their reconstruction based on dynamic bit allocation has been successfully implemented as well, showing no significant loss of information.