next up previous [pdf]

Next: Field data test Up: Synthetic data test Previous: 2D poststack model

3D prestack French model

We created a 3D example by Kirchhoff modeling (Figure 9a), the corresponding velocity model is a slice out of the benchmark French model (Liu and Fomel, 2010; French, 1974). Three surfaces of Figure 9a illustrate the corresponding slices at time=0.6s, midpoint=1.0km, and half-offset=0.2km. Figure 9b is the noisy data. The challenge in this example is to account for multi-dimension, nonstationarity, and conflicting dips. For comparison, we use 3D $ f$ -$ x$ -$ y$ NRNA (Liu and Chen, 2013) to attenuate random noise (Figure 10a). The 3D NRNA method produces a reasonable result. In Figure 10b, 3D $ f$ -$ x$ -$ y$ NRNA shows a better ability of signal preservation than 2D version (Liu et al., 2012). However, 3D NRNA still produces artificial events parallel with curved events. We design a 3D $ t$ -$ x$ -$ y$ APF with 5 (time) $ \times$ 4 (midpoint) $ \times$ 4 (half offset) coefficients for each sample and 15-sample (time), 10-sample (midpoint), and 10-sample (half offset) smoothing radius to further handle the variability of event. Figure 10c and 10d show the denoised result and the difference between noisy data (Figure 9b) and the denoised result (Figure 10c) plotted at the same clip value, respectively. The proposed method succeeds in the sense that it is hard to distinguish the curved and conflicting events in the removed noise. Meanwhile, the 3D $ t$ -$ x$ -$ y$ APF has fewer artificial events, which are more visible for 3D $ f$ -$ x$ -$ y$ NRNA (Figure 10a).

fdata fnoise
fdata,fnoise
Figure 9.
3D synthetic data (a) and the corresponding noisy data (b).
[pdf] [pdf] [png] [png] [scons]

tpre diff3d fpred3 ferr3
tpre,diff3d,fpred3,ferr3
Figure 10.
The denoised result by 3D $ f$ -$ x$ -$ y$ NRNA (a), the noise removed by 3D $ f$ -$ x$ -$ y$ NRNA (b), the denoised result by 3D $ t$ -$ x$ -$ y$ APF (c), and the noise removed by 3D $ t$ -$ x$ -$ y$ APF (d).
[pdf] [pdf] [pdf] [pdf] [png] [png] [png] [png] [scons]


next up previous [pdf]

Next: Field data test Up: Synthetic data test Previous: 2D poststack model

2014-12-07