We have shown that image segmentation with one or multiple attributes can be very effective for 2D seismic data. However, it is in three dimensions that the advantages of automated image segmentation should become even more apparent. While a skilled human interpreter can easily examine a 2D section and pick out a salt interface, visualization and time constraints make this a very difficult process for a 3D survey. In contrast, a computer is not bound by these limitations and can excel at ``seeing" in three dimensions. Furthermore, the drastic increase in the number of pixel-to-pixel comparisons available in three dimensions compared to two should also increase the robustness and accuracy of the segmentation process.
![[pdf]](icons/pdf.png){kind=icon}