Conventional vs Cepstral Based Amplitude and Phase Only Full Waveform Inversion

In earlier work we investigated a cepstrum based approached to full-waveform inversion of seismic data and compared it to the more conventional least squares approach of Lailly (1983) and Tarantola (1984). We observed that it would be theoretically possible to separate the cepstral based approach into amplitude and phase only inversions. In this paper we perform this separation and compare the resulting algorithms to similar conventional amplitude and phase only approaches. We show that in the amplitude only case, conventional amplitude inversion is theoretically stable while cepstral based amplitude only inversion is not. In contrast, we show that cepstral based phase only inversion is not only stable but produces superior results to conventional phase only inversions. Moreover, we show that phase only inversion produces results that are almost as good as those obtained from the full cepstral based method.<br>