Selecting prominent reflectors

 

pixe.xzkm
Figure 9 The reflector points used in the velocity estimation algorithm. The objective function is obtained by summing the semblance at the appropriate value of $\gamma$ for each of these reflector points.

The dips of events in the stacked section are needed for the computation of ${\bf G}$ and can be obtained by computing local dip spectra and picking the maxima. The velocity information of nearby reflector points with the same dip is redundant especially when the velocity varies smoothly laterally. Selecting a sparse set of reflector points to use in the computation of the objective function reduces the cost of the inversion. The reflector points that were chosen were locally maxima in dip semblance and had high minimum entropy norms (Berkhout, 1984) of their dip spectra. It is especially worthwhile to try to select points sparsely normal to the dominant dip and enhance continuity along the dip direction. The picking algorithm penalizes nearby picks in the direction normal to the dominant dip and encourages picks that form continuous prominent reflectors. For the first run of the inversion I also chose not to pick shallow events. Figure 9 shows the reflector points that were picked and used for updating the interval slowness model.