Conclusions and future work

Our modified segmentation method successfully tracked the salt boundaries in all of our test cases. These test cases presented challenges that include a salt boundary that was not continuous and a field data example that had variable amplitude.

Application of this method to 3D datasets is going to be a computational challenge. The normalized cut method has already been applied to paint 3D reservoir models based on coherency data Hale and Emanuel (2003). Unfortunately, 3D seismic cubes can have many more data points. However, for our application of tracking the salt boundary, much of the data can be windowed away in the beginning, hopefully making it possible.