Summary

Coherence attributes are a useful tool for detecting various geological features such as faults/fractures, channels, salt domes etc. in seismic volumes. A problem with interpretation of coherence volumes is that subtle changes in coherence pertaining to the regions of low seismic amplitude, lower levels of discontinuity, and noise are not discernible by the naked eye. Moreover, the regions of multiple discontinuities often appear to be smeared thus causing difficulties in interpretation. The features of interest in the coherence volumes exhibit low coherence values. We present an attribute which enhances low coherence regions using the eigen-decomposition of the Hessian matrix. The resultant volume delineates subtle trends in the coherence volumes which are not visible directly. Moreover, the discontinuities in the highly complex regions are also better localized, thus presenting the interpreter with a greater level of detail about the seismic data.

Introduction

Seismic coherence attributes are often consulted by the interpreters to map various structural and stratigraphic features such as faults and channels. These attributes are a direct indicator of the changes in seismic reflector characteristics and contain a huge amount of information which is not easily interpreted by visualising the seismic data directly.

A problem associated with the coherence attributes is that they do not exhibit a sharp contrast in the regions where the reflector strength or the amount of discontinuity itself, is low. These regions are not easily detectable unless a post-processing operation is performed on the coherence volume.

One of the post-processing algorithms is described by de Rooij and Tingdahl (2002) to achieve enhancement of coherence volumes by iterative searching of azimuths and selection of coherence where the regional difference is found to be maximum. As noted by the authors this method is prone to noise, and parameter dependent for iterative searching.

Another method of enhancing the coherence volume is via 3D Sobel filtering proposed by Chopra et al. (2014) which sharpens the edges of low coherence regions. A pitfall of this approach is that the Sobel operator being a first-order edge detector, detects the edges of non-coherent events instead of the events themselves. This leads to double-responses, one where the low coherence response begins and the other where it ends.

We perform the enhancement of low coherence regions by utilizing an analytical approach based on the Hessian matrix of the coherence volume adapted from the methods mentioned by Oliver (1996) for medical image analysis. The benefits of this approach are fewer parameters, robustness in the face of noisy data, and isotropic behaviour as there is no dependency on search directions. The approach is computationally inexpensive, requiring 1st and 2nd order derivatives in only three orthogonal directions.

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