Abstract:

As Synthetic Aperture Radar (SAR) technology advances and the resolution and quality of SAR systems improves, there is an increasing need for lightweight compression of SAR raw data. In most satellite-borne SAR systems, due to their limited processing capacity, raw data needs to be transmitted to a ground station for processing. As the resolution and acquisition quality increases, so does the volume of data to be transmitted, making compression necessary. Furthermore, computational constraints on-board such systems impose sever restrictions on the kinds of algorithms that can be implemented, and, therefore, on the compression quality. This report proposes a novel lightweight compression approach, based on the principles of universal quantization, which allows the compression system to exploit the structure of the signal in hindsight, i.e., during the decompression stage. This approach shifts the computational complexity to the decoder, which needs to impose the appropriate image model to recover the data. Thus, the heavy lifting in this approach is performed by the decoder at the ground station, which has significantly more computational resources available.