next up previous print clean
Next: Conclusion Up: Lomask and Fomel: Flattening Previous: Computational cost savings

results

We compared the cosine transform algorithm to the FFT algorithm using the synthetic with simple dipping planes displayed in Figure [*]. Although it is just a simple model, it still requires mirroring boundary conditions for the FFT method because the divergence of the dip ($\nabla\it{^T}{\bf r}$) in equation (5) is not periodic. The result of one iteration of FFT flattening with mirrors is displayed in Figure [*]. Notice it is perfectly flat. If we apply the same FFT algorithm without mirrors, we get the result displayed in Figure [*]. It is clearly not flat. If we apply the DCT method as shown in Figure [*], it is flattened in one iteration using only a fraction of the memory and computations.

 
plane3D.fft_no_mirr.flat_cos
Figure 3
The data in Figure [*] flattened using the FFT method without mirrors.
plane3D.fft_no_mirr.flat_cos
view burn build edit restore

 
plane3D.cos.flat_cos
Figure 4
The data in Figure [*] flattened using the DCT method without mirrors.
plane3D.cos.flat_cos
view burn build edit restore


next up previous print clean
Next: Conclusion Up: Lomask and Fomel: Flattening Previous: Computational cost savings
Stanford Exploration Project
4/5/2006