First Break - Volume 7, Issue 2, 1989

The slant-stack or T-P domain offers an alternative representation for seismic data. lts theoretical advantages over the conventional time-offset domain have been studied extensively and are well known. In practice, however, the numerical artefacts of the T-P transform, produced by spatial aliasing and aperture limitations, quite often overpower the expected theoretical advantages. Moreover, the presence of random and coherent noise in the data further deteriorates the performance of the transform. Thus, subsequent processing in the T-P domain becomes less effective. In this paper, we present some applications of slant-stack processing for seismic reflection data. The cornerstone is a new improved T-P transform which incorporates a time- and offset-variant velocity filter. The resulting T-P gathers are free of most transform artefacts and exhibit improved signal-to-noise ratios. First, the performance of the transform is shown with applications both to noise-free synthetic data and to field data contaminated with coherent noise. The forward/inverse T-P transform pair is then applied to the common-shot gathers and the results are compared with those obtained from conventional frequency-wavenumber (f-k) filtering. FinaIly, we combine the forward transform with elliptical moveout (EMO) and stacking and obtain stacked sections in the slant-stack domain. This processing sequence avoids the need for an inverse transform and has several advantages over conventional norm al moveout (NMO) and stacking in the time-offset domain.

Noise from ground roll on common-shot records can be suppressed by the application of windowing in both the time-offset (t-x) domain and the frequency-wavenumber (f-k) domain. The shot record is split into two regions by a reject window which isolates the ground roll and an accept window which selects the remaining data. The data in the reject region are filtered by suitable windowing in the f-k domain. The process is completed by summing the data from the two regions back together. Windowing noise in t-x prior to f-k filtering provides two distinct advantages when compared to filtering the whole record: only those data contaminated by the noise are filtered, and the f-k filter tends to perform better on the restricted data region. The improved performance, although dependent upon filter design and implementation, is due largely to increased zero-padding. Field data examples show improved character of the stack section when only the data in the reject windoware filtered. Ground roll, that propagates on a direct path from source to receivers, gives rise to noise which occupies a fan on seismic records. The reject region is defined to include this noise fan. Synthetic and field data show that event continuity is preserved across the boundary of this reject region after windowing the noise in both domains. Distortion from edge effects is reduced by prior windowing in t-x, because data outside the reject window are available during f-k filtering, while only filtered data inside the reject region are retained.