Introduction

Image segmentation Hale and Emanuel (2002, 2003); Shi and Malik (2000) for tracking salt boundaries Lomask et al. (2004); Lomask and Biondi (2003) is extremely memory intensive. Memory saving measures must be implemented in order to consider applying this technique to 3D seismic cubes. If coarse bounds can be picked, either manually or using another automatic algorithm, this image segmentation algorithm can then be used to partition between the bounds. Unfortunately, the quality of the segmentation result is strongly affected by the shape of the image. For example, elongated images are more likely to be partitioned along their shortest dimension.

In this note, we present one such memory saving technique and demonstrate its ability to pick a salt boundary on a 2D seismic section. By imposing bounds, we significantly reduce the size of the problem and, as a result, increase efficiency and robustness. Also, errors created by segmenting thin images can be rectified with novel boundary conditions described here.