Compression of tactical real-valued SAR imagery in the complex SAR phase history domain

This paper introduces a technique that exploits the statistical properties of complex synthetic aperture radar (SAR) data to effectively compress the information required to produce tactical real-valued SAR imagery. Performing the compression on the complex data as opposed to the real-valued image enables the inherent sinusoidal structure of the radar returns to be modeled. The compression technique also takes advantage of the fact that the phase information in the complex SAR data is discarded in the image formation process. This enables a vocoding technique where white noise is passed through the source model to effectively recreate the frequencies required to reconstruct the SAR image. Line spectral pairs are used to quantized the model parameters and a 45:1 compression ratio is achieved with very little loss in perceptual image quality compared to the JPEG standard.