next up previous [pdf]

Next: Conclusions Up: Chen et al.: Deblending Previous: Median filtering after NMO

Examples

cmpa cmp semblancescn
cmpa,cmp,semblancescn
Figure 6.
(a) Clean CMP gather. (b) Blended CMP gather. (c) Velocity scan for blended CMP gather.
[pdf] [pdf] [pdf] [png] [png] [png] [scons]

semblancenmo semblancenmomf
semblancenmo,semblancenmomf
Figure 7.
(a) CMP gather after NMO. (b) CMP gather after NMO and median filtering.
[pdf] [pdf] [png] [png] [scons]

semblanceinmo semblancediffmf semblanceerrormf
semblanceinmo,semblancediffmf,semblanceerrormf
Figure 8.
(a) CMP gathering after deblending. (b) Blending noise section. (c) Deblending error section.
[pdf] [pdf] [pdf] [png] [png] [png] [scons]

The first example is a single synthetic CMP gather. The clean unblended data, blended data, and velocity scan of the blended data are shown in Figures 6a, 6b, and 6c, respectively. Figure 7 shows the blended data after NMO correction and deblended data after NMO correction and median filtering. The median filtering is effective in that most of the interferences have been removed. After inverse NMO on Figure 7b, the deblended data in CMP gather is shown in Figure 8a. The blending noise section is shown in Figure 8b. From the deblending error section as shown in Figure 8c, we conclude that the proposed method can achieve an excellent result, because the deblending error is small.

data data-b data-db
data,data-b,data-db
Figure 9.
Comparison of CMP gathers. (a) Original unblended CMP gathers. (b) Blended CMP gathers. (c) Deblended CMP gathers.
[pdf] [pdf] [pdf] [png] [png] [png] [scons]

svscan svscan-b svscan-db
svscan,svscan-b,svscan-db
Figure 10.
Comparison of velocity scans. (a) Velocity scan for original unblended data. (b) Velocity scan for blended data. (c) Velocity scan for deblended data.
[pdf] [pdf] [pdf] [png] [png] [png] [scons]

spstm spstm-b spstm-db
spstm,spstm-b,spstm-db
Figure 11.
Comparison of migrated images using prestack Kirchhoff time migration (PSKTM). (a) Migrated image for unblended data. (b) Migrated image for blended data. (c) Migrated image for deblended data.
[pdf] [pdf] [pdf] [png] [png] [png] [scons]

We now provide two examples to demonstrate the performance of the proposed workflow described previously. In the next two examples, we use two sources to simulate the blended data. The second example is based on a simple synthetic dataset, which contains four reflectors. The velocity in the model is linearly increasing along the depth axis. We use Kirchhoff modeling to simulate the CSG and blend the data according to IMSSS. After common velocity-semblance scanning, we can pick the NMO velocity and apply NMO. Applying median filtering with a 9-point filter length along the offset direction removes the blending noise. By inverse NMO, we get the deblended dataset in the common-midpoint domain. Figure 9 shows the comparison of CMGs. Figure 10 shows the comparison of velocity scan, in which we see the velocity scan for the blended CMG is smeared along with the noise. After scanning the deblended data coming from the first rough velocity scan and first NMO correction, we however obtain a convincing velocity map. Using the updated NMO with new velocities, we get flatter events that are more suitable for median filtering. Figure 11 shows the comparison of migrated images. In this example, we use prestack kirchhoff time migration (PSKTM) (Docherty, 1991) as the migration operator. We then use Dix inversion (Dix, 1955) to convert the time image to depth image. The migrated image after deblending is much cleaner than that of the blended data.

gulf gulf-b cmps-db
gulf,gulf-b,cmps-db
Figure 12.
Comparison of CMP gathers. (a) Original unblended CMP gathers. (b) Blended CMP gathers. (c) Deblended CMP gathers.
[pdf] [pdf] [pdf] [png] [png] [png] [scons]

vscan-gulf vscan-b vscan-db
vscan-gulf,vscan-b,vscan-db
Figure 13.
Comparison of velocity scans. (a) Velocity scan for original unblended data. (b) Velocity scan for blended data. (c) Velocity scan for deblended data.
[pdf] [pdf] [pdf] [png] [png] [png] [scons]

pstm pstm-b pstm-db
pstm,pstm-b,pstm-db
Figure 14.
Comparison of migrated images using prestack Kirchhoff time migration (PSKTM). (a) Migrated image for unblended data. (b) Migrated image for blended data. (c) Migrated image for deblended data.
[pdf] [pdf] [pdf] [png] [png] [png] [scons]

pstmzoom pstm-bzoom pstm-dbzoom
pstmzoom,pstm-bzoom,pstm-dbzoom
Figure 15.
Comparison of zoomed migrated images using prestack Kirchhoff time migration (PSKTM). (a) Zoomed migrated image for unblended data. (b) Zoomed migrated image for blended data. (c) Zoomed migrated image for deblended data.
[pdf] [pdf] [pdf] [png] [png] [png] [scons]

The third example is a marine field dataset from the Gulf of Mexico. Figure 12 shows the comparison of CMGs. Figure 13 shows the comparison of velocity scans. Figure 14 shows the comparison of the migrated image. In this example, we use the same migration approach to obtain the seismic images. We have the similar observation to that for the second example: the migrated image for the deblended data is cleaner, especially for the shallow part, indicated by the arrows and frame boxes. For a better view, we zoom the parts indicated by frame boxes and show them in Figure 15. It's obvious to see the improvement for the final migrated image after using the proposed deblending approach.


next up previous [pdf]

Next: Conclusions Up: Chen et al.: Deblending Previous: Median filtering after NMO

2014-11-10