Automatic interpretation of salt bodies by iterative image segmentation

Velocity model building is the most human-intensive component of the depth-imaging<br>process, and it is often the bottleneck when trying to reduce the cycle time of large seismic<br>imaging projects. For near-salt or sub-salt imaging the interpretation of the salt-body<br>geometries can be extremely time consuming. Current automatic methods based on horizon<br>tracking are prone to errors, in particular when the salt boundaries are poorly imaged. Lomask<br>et al. (2007) have proposed an automatic method to interpret salt boundaries that segments the<br>image cube by solving a global optimization problem, and thus it is more robust than local<br>methods based on horizon tracking. We apply the image-segmentation method to the iterative<br>velocity-model building process. We show how it can be applied to a conventional sediment<br>and salt flooding procedure and we discuss how to use the boundaries picked at the previous<br>iterations as a constraint to the iterative solution and thus make the method more reliable.