Compressive Sensing vs. JPEG: An End-to-End Comparison

We present an end-to-end image compression system based on compressive sensing by adding measurement quantization and entropy coding to the conventional scheme of compressive sampling and reconstruction. The compression performance, in terms of data rate versus image quality, is shown to be on par with JPEG and significantly better at the low rate range. We study the parameters that influence the system performance, including the choice of sensing matrix and the trade-off between quantization and compression ratio, and we provide an effective method to control both them jointly in order get near optimal quality at any given bit rate.