The Study Of Data Compression Algorithm For Synthetic Aperture Radar At Low Bit Rates

This dissertation focuses on the study of the compression algorithms of Synthetic Aperture Radar(SAR) raw data and imagery at low bit rates. For the SAR raw data, the dissertation mainly discusses data compression algorithms in the transform domain, such as, algorithms based on the lifting wavelet transform or discrete cosine transform(DCT). For the lossy compression of SAR imagery, the dissertation mainly discusses compression algorithms based on the combination of wavelet transform and zero-tree quantization. For the lossless coding of SAR imagery. it mainly discusses coding algorithms based on bit-plane coding.The dissertation analyzes the source model of SAR raw data and imagery in detail and the baseline theories of quantization and coding are discussed. The theories of 2-band wavelet transform are deeply studied, including: theories of multi-resolution analysis , fast wavelet transform and lifting wavelet transform. Based on the discussion of baseline theories of data compression, new algorithms and improved versions of origin algorithms are presented.The major devotement of this dissertation is as follows:1. A new algorithm for compressing SAR raw data using block lifting wavelet transform is presented. In this algorithm, problems of wavelet basis function selection, data edge processing and wavelet decomposition levels selection are discussed in detail and the statistical properties of wavelet coefficients are deeply analyzed. Due to the statistical properties of SAR signal, a proper sub-band bit allocation strategy is proposed. The performances achieved in terms of bit rate reduction and certain quality parameters have been evaluated. These show an increased performance in this method with respect to conventional methods.2. Based on the discussion of a fast two-dimensional DCT algorithm implementing with one-dimensional DCT, a compression algorithm for SAR raw data based on the combination of DCT and block adaptive quantization(BAQ) is presented. This dissertation exploits the known results that a block normalized SAR raw signal is a Gaussian stationary process to re-array the DCT coefficients. Coupled with a proper bit allocation strategy and block adaptive quantization on re-arrayed coefficients matrix,...