Abstract
Buried channels are considered stratigraphic traps, essential in petroleum exploration and drilling hazard management. A thorough investigation of the buried channels provides helpful information about the sedimentation processes and the marine currents. Hand-operated interpretation of channels is sometimes laborintensive and time-consuming, particularly in areas with complex geological patterns or low signal-to-noise ratios. However, automatic feature extraction is a task-on-demand. Channel boundaries, depicted as curvy and curvilinear events on time slices, are called edges in image processing, where amplitudes intensively vary. This paper provides an improved edge delineation algorithm with a sub-pixel resolution for an enhanced channel boundary depiction. The partial area effect, in combination with a unique edge linking technique, is proposed to strengthen the connectivity of edge segments and the fidelity of edge outlines. First, a gradient-directed partial area effect mask was employed in the edge area of interest to extract the edge position in each pixel with relatively high precision and less affected by environmental noise. Then, an edge linking algorithm based on the edge segments' distance was applied to the partial area effect's results to connect the circular arc edges detected. Field tests were carried out on synthetic and field data sets containing several channelized features with differing widths and tortuosity. The proposed sub-pixel procedure afforded more precise and cost-effective outcomes than approaches with pixel resolutions. The quantitative validation tests using root mean square error, peak signal-tonoise ratio, structural similarity index, and Pearson correlation coefficient revealed that the proposed approach exceeded the traditional edge measures and seismic attributes by detecting more reliable edge points. Finally, we tested the proposed algorithm on synthetic and field seismic data sets containing salt domes and fault events to determine its applicability for localizing other geological features. The detected edges were quantitatively validated based on the manually interpreted events. The salt and fault boundaries detected by the proposed algorithm have relatively high coincidences with their ground truths.
NSTL
Elsevier