Automatic seismic horizon extraction via optimal growing with a priority queue and global fitting with multisource information

ABSTRACT Horizon extraction is fundamental in seismic interpretation but can be time consuming when done manually. Although many methods have been developed to automate horizon extraction, most still require significant human intervention, especially for extracting horizons correctly crossing faults, stopping at intrusions and unconformities, and consistently following seismic phases. To address these challenges, we develop a fully automated horizon extraction algorithm that first performs optimal growth using a priority queue, followed by global fitting with multisource information. We begin by identifying all amplitude peaks or troughs as candidate points for horizon growth, characterized by attributes such as amplitude, slope, planarity, and waveform segments. The key to horizon growth is finding an optimal path to link points likely belonging to the same horizon. This is achieved using a priority queue that continuously collects candidates and automatically sorts them, selecting the highest-priority point first for horizon growth. With the selected point, its nearby points within a 3D window are identified as candidates and added to the priority queue if their attributes are sufficiently similar to those of the selected point, according to specific measurements. These measurement values also are used as priority indicators for sorting. This iterative process continues until the queue is empty, completing the horizon growth. By selecting the highest-priority candidate to search for nearby points at each step, the growth process navigates around faults, halts at intrusions and unconformities, and avoids low-quality areas to track a geologically reasonable horizon. However, the grown horizon may exhibit holes in areas of low data quality. To address this, we further compute a least-squares fit of the point positions of the grown horizon, seismic slopes, and multigrid correlations to produce a more complete horizon. We demonstrate the effectiveness of our method on multiple field seismic volumes complicated by faults, intrusions, unconformities, and noise.