Fourier methods of seismic data regularisation

Conclusions

This paper has shown that Fourier methods of seismic data regularisation can be a powerful and robust method in regularising data, which is useful for future processing steps and for image viewing. A standard irregular DFT has limited application since the contaminated spectra lead to a large number of artifacts and amplitudes are not reconstructed correctly. Using a recursive ALFT approach gives a much improved result - amplitudes are now correct along events, events are exactly aligned in the correct places and there is a slight reduction in artifacts. Discontinuous events are imaged and whilst aliased events are not perfectly reconstructed they do not contaminate the image as strongly as in the standard DFT case, and this method can be extended to handle these sorts of frequencies. However the ALFT approach is slow, and could be improved by implementing a non-uniform FFT and better tapering between transforms, to help reduce artifacts.

Examination of multitaper methods with Slepian tapers has yielded some intriguing results - the spectra after tapering and stacking are shown to have reduced spectral leakage without iterations around frequencies of interest, however leakage noise at higher frequencies is preserved.

Key issues that still require resolving before this is proved to be a viable regularisation method are those of adapting taper weights to optimise the amplitude spectrum and not the power, a better method of reconstructing and weighting the phase spectrum and a more robust method of removing the taper imprints in the reconstruced data. Early tests are showing that these problems are gradually being overcome.

Fourier methods of seismic data regularisation