Hale and Emanuel (2002, 2003) applied the normalized cut image segmentation method developed by Shi and Malik (2000) to paint a coherency based reservoir model. Our approach is very similar. This image segmentation technique creates a matrix containing weights relating each pixel to every other pixel in a local neighborhood. The matrix is then used to cut the image where the normalized sum of weights cut is minimized. We have modified the weight calculation to be inversely proportional to the absolute value of the complex trace (instantaneous amplitude) of the seismic. This makes the weights very weak at salt boundaries, causing the segmentation algorithm to cut along the boundary.
In this paper, we give a very general overview of the normalized cut segmentation technique. We then describe how we modified it for application to salt dome seismic data. We test this technique on simple models illustrating its efficacy with discontinuous salt boundaries. We also apply this method to a 2D field section where the amplitude is inconsistent and challenging to pick. Methods for dealing with noisy data are also presented.