Block Floating Point Quantization for RAW and Processed Synthetic Aperture Radar Data

Abstract

Synthetic Aperture Radar (SAR) is an imaging radar, primarily used to generatemicrowave images of earth surface. SAR images nd applications in various areas likeWater resources' management, Disaster management, Agriculture etc. SAR generateshuge volumes of raw data with very high data rates. This places severe demands onthe onboard Solid State Recorder storage as well as spacecraft to earth station down-link. This in turn restricts the extent of SAR instrument operation and consequentlythe global coverage. Onboard raw or processed data compression of raw SAR data isthus an essential requirement for SAR sensor because of limited onboard data storageand limited downlink bandwidth.Block Adaptive Quantization (BAQ) is used for SAR data compression. BAQconsists of adapting the step size of a quantizer based on mean and variance of theinput and then quantize each sample to the respected level of quantization.In this dissertation, an algorithm is proposed which can be used along with theBAQ for compression of SAR data. The proposed algorithm is an iterative algo-rithm, designed to give optimum SNR results with desired compression amount. Atinitial stage, algorithm calculates both reconstruction and threshold levels by a simplemethod. Then in next stage it calculates reconstruction levels with reference to pre-vious thresholds and based on those reconstruction levels, thresholds are calculated.In this report, Matlab simulation results of proposed algorithm are given. Theresults include various compression modes for various input distributions like Gaus-sian, Laplacian and Sinusoidal. Further the results achieved are also compared withthe methods currently implemented at SAC, ISRO in terms of SNR. Xilinx synthesisresults for xed 3 bit compression mode are also shown.