Nonstationary filtering techniques can be used to create nonstationary deconvolution operators designed directly from the seismic data and apply them to the data. Such operators can be continuously time-variant and have any desired amplitude and phasespectra. The operator design uses time-variant Fourier or Burg spectra measured directly from seismic data, which are smoothed, inverted, and combined with a minimum-phase spectrum, if desired. This method of deconvolution, named time variant spectral inversion (TVSI), approximately corrects the seismic data for the effects of anelastic attenuation, frequency dispersion, and source signature. The result is a onedimensional nonstationary operation which extends the range of stationary deconvolution to a type of data-driven inverse-Q filter.
View full article as PDF (0.61 Mb)