Multi Channel Operations Two Dimensional Filters
Of all the tools which are available for processing seismic data the technique of band-pass filtering is the oldest. It is still useful in some cases although current good practice is to acquire and preserve the broadest possible frequency band.
A properly selected digital band-pass filter can eliminate noise very effectively when the frequency of the noise differs from that of the signal. But today most noise lies within the wider range of signal frequencies. The problem is now one of degree expressed by the relative signal-to-noise ratio. In all such cases the noise cannot be removed by band-pass filters without removing some signal also.
Deconvolution does not, as a rule, remove noise. Spiking deconvolution merely restores weak frequency components to their proper level. But if the restored components are noisy their influence on the output will increase and they will degrade rather than improve the output.
Conversely, deconvolution may also suppress highly colored (very strong) spectrum components. This tends to eliminate oscillatory noise from the signal by merely suppressing the cyclic component without restoring the correct parameters for that frequency. Most likely the phase component will be incorrect.
Predictive deconvolution may be used to eliminate multiple reflections from the seismic signal when the signal meets the assumptions of the model upon which the deconvolution method is based.
Most seismic noise elements have at least some of the same frequency content as the desired signal and, in most cases, are extremely difficult to remove by any standard single-trace filter method.
Figures & Tables
“This publication originated in 1967 as a few notes to accompany a basic seminar for the Canadian SEG and was expanded in 1968 into an SEG Continuing Education course. Old and new information about geophysical data processing is consolidated in this edition. How to choose processes and parameters for any given field data is shown.”