Evaluation of differentiation for lossless compression of several signal classes

For some classes of signals, particularly those dominated by low frequency components, such as seismic data, first and higher order differences between adjacent signal samples are much smaller compared with the signal samples. In this paper, evaluating the differencing approach for losslessly compressing several candidate classes of signals is given. Three different approaches employing derivatives are applied yielding different results for each signal class. The performance of these lossless techniques and the adaptive linear predictor are compared for the different signal categories. Preliminary results indicate that the derivative technique, of different orders, for lossless compression is promising. Although the residual may be slightly higher compared with the linear predictor approach, the differentiator coefficients are integers which do not have to be encoded and much fewer computations are needed to compute the residual.