Automatic Horizon Picking in 3D Seismic Data Using Optical Filters and Minimum Spanning Tree (Patent Pending)

Horizon picking in 3D seismic data is a very challenging problem. The difficulty for automatic horizon extraction exists at least in two fold: (1) the selection of picks in a trace usually ignores lateral continuity, and (2) the trace traversal order can result in significantly different horizons so that the resulting picks in the same horizon often conflict with each other. In this paper, a pattern recognition-based algorithm is presented to explicitly address these two difficulties: (1) select a pick within a trace by considering context information through orientation filters which help to preserve the lateral continuity among traces; (2) perform the trace selection using the minimum-spanning tree (MST) algorithm based on the confidence maximum at each pick. Combining the pick selection and trace selection components together allows us to obtain highly accurate horizon surfaces.