Examples

Stable estimate of dips in the stacked zero-offset or migrated section are required in order to make super-gather corrections. Typical results of such dip scans are shown for an inline of the SEG/EAGE Salt model (Figure ). Claerbout (1992) introduces plane-wave deconstruction filters that locally decompose an image into plane waves, and Fomel (2002) extends the concept to a least-squares method for continuously decomposing an image into dips. Figure () shows the result of applying this algorithm to the 2D stack image. A semblance scan provides an alternative--and somewhat simpler--method for computing dips. This method is identical to CRS post-stack parameter estimation Jäger and Hubral (2001). At evenly spaced locations in the stack, a grid search maximizes semblance across circular slices through the volume for different values of the dip $\mathbf{\sigma}$. For 3D data, the search iterates over both x- and y-components of the dip, so it can be expensive to finely sample the search space. A staged search strategy can help overcome this problem by first determining coarse dip estimates and then iteratively refining the results with finer search grids. Figure () shows the dip section from the semblance scan.

 

seg_inline
Figure 2 An inline from a zero-offset stack of the SEG/EAGE Salt model.

 

fomel_dips
Figure 3 Dips computed for the SEG/EAGE Salt model after Fomel (2002).

 

semb_dips
Figure 4 Dips computed for the SEG/EAGE Salt model using a semblance scan.

 

seg_vel_pan
Figure 5 Velocity panel for a 5x5 super-gather computed without dip corrections (left) and with dip corrections (right). The dip corrections improve the coherence of the semblance peaks and make the panel easier to pick.

Figure () shows velocity spectra computed on a super-gather from the salt model. The gather contains data from 5 inlines and 5 crosslines, and the spectra are computed with NMO (left) and with the dip corrected move-out (right). While this example shows that dip corrections can improve the quality of semblance panels for super-gathers, it does not motivate the need for super-gathers. Since the salt model does not contain noise, a perfectly adequate velocity spectrum can be computed from a single CDP. But in real data, especially land datasets, noise is a major impediment for velocity analysis. The next example tests the correction on a 3D land dataset. The data were collected with 25m spacing in both the in-line and cross-line directions in a region with complex geology characterized by salt intrusions. With an average fold of 25, noise is a significant problem for velocity analysis.

 

• Land example