3-D missing data interpolation

 

qdome
Figure 27 Claerbout's ``qdome'' synthetic model. Left: original model. Right: input to interpolation (60% of the traces removed).

 

qslope
Figure 28 Plane-wave slope estimates in the x and y directions (left and right plots, respectively) from the ``qdome'' model. Top: estimated from the original model. Bottom: estimated from the decimated model.

Figure  shows Claerbout's ``qdome'' synthetic model Claerbout (1993, 1999), which models a seismic image of a complicated sedimentary geology. In a data regularization experiment, I randomly remove 60% of the traces in the original model, arriving at the missing data model shown in the right plot of Figure . The 3-D slope estimates from the input data are shown in the bottom plots of Figure . The estimates fairly accurately match the slope estimates obtained from the original model (top plots in Figure ). The missing-data interpolation result is shown in Figure . Most of the original signal has been successfully restored. Some of the fault sharpness is lost in the interpolation result, but all the curved and planar events are accurately preserved. Clapp (2000a) reports a successful interpolation result using a more severe decimation of the same synthetic model. However, he assumes a prior knowledge of the local dip field instead of estimating the dips from the decimated data.

 

pmiss Figure 29 Result of missing-data interpolation with a 3-D local plane-wave destruction filter. Compare with Figure .