Abstract: A seismic data compression technique is disclosed which comprises sampling each individual seismic trace, operating upon a set number of samples to generate a predicted sample and quantizing the difference between the next sample and the predicted value of the sample, and transmitting the quantum number whereby the amount of information which need be transmitted is limited. In a preferred embodiment, a linear prediction differential pulse code modulation scheme is used to provide the predicted value, while an adaptive quantization scheme is used to quantize the error value to be transmitted, thus yielding further improvements in accuracy. A feedback loop can be applied to the decompression operation to limit quantization noise and further improve the fidelity of representation of the decompressed signals.