Edge-preserving smoothing for segmentation of seismic images

Field Data Examples

Figure 6 shows a 3D field data image from the Gulf of Mexico. The salt boundary is clearly visible, but its discontinuous nature can pose problems for many automatic picking algorithms. Figure 7 shows the result of smoothing the image with traditional filtering (a), and the hybrid-MH method (b). The hybrid-MH filter attenuates a noticeable amount of background noise, while preserving the salt boundaries much more clearly than traditional smoothing. The difference calculations shown in Figure 8 confirm that MH-filtering removes much less coherent signal than traditional smoothing.

img-orig
Figure 6. A 3D image from the Gulf of Mexico with a prominent but discontinuous salt boundary.

Figure 9 shows the effects of varying the hybrid threshold parameter $\alpha$ for the field data example. As $\alpha$ decreases, the image becomes noticeably smoother, even though the reflector boundaries remain clear. In this example, though, it is evident that the reflectors' amplitudes are being modified, which is an important consideration.

img-smth,img-hyb-1
Figure 7. The image in Figure 6, smoothed with (a) a traditional box filter, and (b) a hybrid-MH filter of the same length.

diff-smth,diff-mhn
Figure 8. Results of differencing the original image in Figure 6 with (a) the image smoothed with a traditional box filter, and (b) the image smoothed with the MH filter. Again, the result in (b) shows that much less coherent signal has been removed from the image.

img-hyb-50,img-hyb-20,img-hyb-10,img-hyb-01
Figure 9. Results of applying the hybrid-MH filter on the image in Figure 6, with $\alpha$ set at (a) 0.5, (b) 0.2, and (c) 0.1 and (d) 0.01. While the reflector ampliltudes are affected, a great deal of speckle noise is removed at low $\alpha$ values.

Edge-preserving smoothing for segmentation of seismic images