Summary

Fault picking is a critical, but resource intensive tool for seismic interpretation. In a bid to improve fault imaging in seismic data, we have applied a directional Laplacian of a Gaussian (LoG) operator to sharpen fault features within a dissimilarity volume. We compute an M by M matrix of the distance-weighted coherence values that fall within a 3D analysis window about each voxel. The eigenvectors of this matrix define the orientation of planar discontinuities while the corresponding eigenvalues determine whether these discontinuities are significant. If a plane is detected, the eigenvectors define a natural coordinate system for smoothing. We rotate the data to the new coordinate system and apply the sharpening operator. We find that our proposed algorithm improves the faults image by sharpening such features and suppressing some of the noise within. The reliability and robustness of the technique is tested on real data from New Zealand.

Introduction

Fault image enhancement remains a pivotal objective in 3D seismic interpretation. Interpreting fault features is time consuming and may sometimes present a great deal of unease. For this reason, initiatives geared towards minimizing this intense human labor can be incredibly valuable.

The application of image processing techniques for fault enhancement has garnered much attention in the geophysics community. In 2003, AlBinHassan and Marfurt evaluated the Hough transform as a fault enhancement algorithm. Aare and Wallet (2011) and Dorn and Kadlec (2011) show considerable improvements on this approach that better define the continuity of larger faults. Donais et al. (2007) address a fault attribute based on a robust directional scheme. This technique is well-suited to differentiate faults from stratigraphic features and consists of performing an eigenstructure analysis of the gradient vector field covariance matrix along a segment. Their interpretation states that groups of disorder patches are more likely to represent faults while more isolated patches are more likely to represent stratigraphic features. Lavialle et al. (2007) introduce a nonlinear filtering technique based on the definition of a partial differential equation that de-noises and preserves faults before automatic fault extraction. In his work, Barnes (2006) suggests using precomputed coherence data to define potential fault anomalies. He then defines a filter to pass near vertical planar coherence anomalies. These anomalies are further joined through dilation and sharpened by skeletonization to provide a fault overlay on the amplitude data volume.

This content is only available via PDF.
You can access this article if you purchase or spend a download.