Seismic data compression using adaptive filters

This paper describes a technique for waveform data compression. The technique is a variation on a two-stage lossless method that was developed by one of the authors. Whereas the earlier work employed a special version of the classical linear predictor in its first stage, the present method uses a lossless successive difference operation combined with an adaptive differential pulse code modulation-type (ADPCM) scheme to output a integer residual sequence for transmission, storage, or further encoding. The present method was found to operate best by taking the fourth difference of the data. Two ADPCM implementations are considered; the basic FIR filter scheme and a frequency selective scheme proposed earlier. To reconstruct the original integer data, only the first four unfiltered data values and the initial adaptive filter parameters are needed in addition to the integer residual sequence. The proposed adaptive scheme is compared to that of the linear predictor for seismic data. Results show that the original integer sequence can be recovered without loss