Vector quantization for saturated SAR raw data compression Original Research Article

Spaceborne SAR involves the storage and transmission of large-size sampling data. Block adaptive quantization (BAQ) is now the most widely used onboard data compression algorithm due to its good tradeoff between system performance and complexity. However, when spaceborne SAR raw data is saturated, the performance of conventional BAQ deteriorates dramatically because its precondition of Gaussian distribution of raw data no longer holds. In order to solve this problem, an improved vector quantization (VQ) algorithm is proposed. This algorithm firstly introduces saturation modification to a conventional vector quantizer, obtains the saturation codebook based on Gaussian density function, and then obtains the new vector quantizer for the whole set of Saturation Degree (SD). This algorithm makes the vector quantizer match statistical model of data for the whole set of SD, so the performance of the compression is improved. The case of the 2D signal is explicitly computed. The performance of the proposed algorithm is verified by simulated and real data experiments.