Enhanced random noise removal by inversion

Noise attenuation by prediction filtering breaks down in the presence of high-amplitude noise when the prediction filter is corrupted by noise and the filter response to the noise overwhelms the signal. Spurious events are generated and the amplitude of the signal is reduced by prediction filtering under these circumstances. To reduce these undesired effects, the separation of signal and noise is posed as an inversion problem. The inversion process preserves signal amplitudes and attenuates spurious events.