next up previous contents
Next: Sensitivity to Velocity Error Up: Numerical results Previous: Plane-wave Prestack LSRTM   Contents

Computational and I/O Cost

Table [*] summaries the computational cost and the I/O cost of each of the migration methods. All the calculations are scaled to the conventional RTM method. The conventional LSRTM in shot-domain is the most expensive method with computational cost 60, assuming each iteration takes twice the computation of one RTM operation and 30 iterations are needed. The computational cost of standard plane-wave RTM is $ \frac{31}{801}=0.04$ and the cost of LSRTM for 30 iterations is $ \frac{31*2*30}{801}=2.32$ times that of standard RTM. The dynamic LSRTM is very efficient and its computational cost is only $ \frac{31*2}{801}=0.08$ of standard RTM. In terms of I/O cost, there are only 31 plane-wave gathers, so the plane-wave RTM and LSRTM methods only have $ \frac{31}{801}=4\%$ of the I/O cost of conventional RTM for the Marmousi2 model example, if the plane-wave gathers can be stored in the memory.


Table 3.1: LSRTM and RTM computational cost, I/O expense, image quality and sensitivity to errors in the migration velocity for the example of Marmousi2 model with a fixed spread acquisition geometry.
  RTM Shot-domain LSRTM Plane-wave RTM Plane-wave LSRTM Dynamic LSRTM
Computation Cost 1 60 0.04 2.32 0.08
I/O Cost 1 1 0.04 0.04 0.04
Image Quality good highest better highest better
Sensitivity to Velocity Errors good fair good good good

         


next up previous contents
Next: Sensitivity to Velocity Error Up: Numerical results Previous: Plane-wave Prestack LSRTM   Contents
Wei Dai 2013-07-10