Design of a data compression scheme for the Canadian National Seismographic Network

The authors propose a novel method for finding optimum coding parameters of bilevel coding or binary run-length coding, which decomposes a multivariable optimization problem into a one-variable optimization, thereby avoiding the associated time inefficiency of trial. It has been shown that, although the improvement of the compression ratio using this coding parameter optimization is negligible, the computation time is more than 10 times faster. Some special considerations for the design of the predictor have been found to be significant for increasing the compression ratio. The proposed approach has been applied to actual geophysical data provided by the Pacific Geoscience Centre of Geological Survey of Canada. The results illustrate that the bits per sample using the proposed techniques are decreased by as much as 1.4 bits compared with ENCODE2; the compression ratio is increased by as much as 24.1% and compress/decompress CPU time is reduced by about 30%