Predictive Quantization of Range-Focused SAR Raw Data

Synthetic aperture radar (SAR) systems create massive amounts of data which require huge resources for transmission or storage. The limited capacity of the downlink channel demands efficient onboard compression of SAR data. However, SAR raw data exhibit very little correlation which can be exploited in a compression algorithm. Range focusing is shown to increase the data correlation by exposing some of the distinctive features of the scene under surveillance. In this paper, we first present analysis of spotlight-mode SAR to show the source of the increased correlation in the range-focused data. Next, we propose two algorithms-transform-domain block predictive quantization (TD-BPQ) and transform-domain block predictive trellis-coded quantization (TD-BPTCQ)-for the compression of the range-focused data. Experimental results indicate that, at the rate of 1 bit/sample, and for similar or lower computational complexity, TD-BPQ and TD-BPTCQ outperform the best method proposed in the literature by 1.5 and 2.3 dB in signal-to-quantization-noise ratio, respectively. Similar improvements are observed for the rate of 2 bits/sample.