Lossless multispectral and hyperspectral image compression on multicore systems

The necessity of performing multispectral and hyperspectral image compression on multicore processor devices running on-board satellites has increased as spectral remote sensing devices produce a high amount of data. Multispectral and hyperspectral instruments acquire images that need compressing before being sent to Earth, because the bandwidth used for transmitting images is a limited resource. This paper presents an effective parallelization scheme of the lossless compression algorithm presented in the CCSDS (Consultative Committee for Space Data Systems) 123.0-B-1 standard. Its implementation exploits the task, data, and pipeline parallelism that exists in the algorithm in order to maximize throughput, reduce execution time, and memory usage. The parallel program is based on the master-worker design paradigm. The implementation results show how the performance of the parallelized algorithm will scale as the number of processor cores increases.