Real-Time Adaptive Lossless Hyperspectral Image Compression using CCSDS on Parallel GPGPU and Multicore Processor Systems

The proposed CCSDS (Consultative Committee for Space Data Systems) Lossless Hyperspectral Image Compression Algorithm was designed to facilitate a fast hardware implementation. This paper analyses that algorithm with regard to available parallelism and describes fast parallel implementations in software for GPGPU and Multicore CPU architectures. We show that careful software implementation, using hardware acceleration in the form of GPGPUs or even just multicore processors, can exceed the performance of existing hardware and software implementations by up to 11x and break the real-time barrier for the first time for a typical test application.