SAR Raw Data Compression In Wavelet Domain

As a high-resolution active microwave imaging radar, Synthetic Aperture Radar (SAR) system generates a big amount of data, which is hard to be transmitted or stored because of limited processing or downlink capacity, so there is an urgent need to reduce the data column. Data compression is an effective method to solve the problem and keep SAR system quota at the same time. After analyzing statistics and power characteristics of SAR raw data, the paper studies data compression of SAR raw data in wavelet domain, from the point of 2-band wavelet transform, 2-band wavelet packet transform, multiband wavelet transform and multiwavelet transform.The paper studies 2-band wavelet transform theory and advances two quantization algorithms according to characterics of the transform coefficients, WT-ECBAQ and WT-TCQ. As the most common method, 2-band wavelet transform analyzes time and frequency information of signals with non-uniform resolution, concentrates energy of transform coefficients, and allocates bit resource to gain good performance. WT-ECBAQ adjusts parameters of quantizer adaptively in line with variance of data block, uses entopy-constrained quantizer to decrease data entropy and gain requested compression ratio. Compression experiments have proved good compression performance. WT-TCQ uses TCQ to quantize wavelet coefficients and gains better compression performance. The advantages of this algorithm are more obvious at large rate.The paper makes comprehensive research into 2-band wavelet packet transform, analyzs the features of transform coefficients and puts forward WPT-BAQ and WPT-TCQ. 2-band wavelet packet transform modulates decomposition structure and selectively decomposes high-frequency component at every level. It conquers the shortcomings of 2-band wavelet transform which has low resolution at highfrequency, and uses code resource effectively by distributing bit ratio to different band. WPT-BAQ tunes quantization parameters of Lloyd-Max quantizer adaptively according to block variances and makes good use of quantization resource. WPT-TCQ obtains good compression performance using limited relativity and goodness of TCQ. Compression experiments show that the performance of WPT-TCQ is better than WPT-BAQ and the difference is much bigger at high rate.The paper quantizes M-band wavelet transform coefficients with BAQ based on the M-band wavelet transform theory. While 2-band wavelet has one scalar function and one wavelet function, M-band wavelet has one scalar function and M-l wavelet functions. So multiband wavelet transform can directly and finely divide signal frequency, which is useful to obtain signal details. The algorithm uses code resource reasonably to decrease quanzation distortion. Compression tests show good performance of this algorithm.Combining properties of SAR raw data and advantages of muhiwavelet, the paper quantizes muhiwavelet transform coefficients with BAQ and analyzes compression performance with experiments.As a new concept, muhiwavelet has vector scalar function and vector wavelet function, and it meets demands for symmetry, compacted support, vanishing moments and regularity needs at the same time.Because of good performance and complexity ratio, SAR raw data compression in wavelet domain is development direction for realtime data compression.