Enhanced Fault Segmentation using an Adaptive 3-D Sobel Filter
Aqrawi, Ahmed A.; Barka, David S.
We suggest an improved method of traversing post-stack seismic data with an adaptive operator size for detecting edges. Attributes, in the coherence family, used for edge detection usually have a static operator size (Marfurt and Chopra, Seismic attributes for prospect identification and reservoir characterization, 2007). This can prove to be limiting in that one would under or over sample in regions of higher or lower frequencies respectively. By looking at the nature of seismic data, which commonly changes from short wavelength to longer wavelength signals in depth/time, an adaptive approach is more suitable for detection of features. By adjusting the operator size in relation to the depth/time, we are better following the frequencies of the data as we filter them. To increase the accuracy and resolution of our filtering we have chosen to interpolate between adjacent seismic traces.
However, seismic data is not that simple in structure such that one can only vary in depth/time. Geological features such as dipping, salt and gas result in chaotic and varying frequencies regardless of when they occur. This is why we have introduced a textural analysis to account for this change and adapt our filtering to it. We have chosen to use chaos (Iske and Randen, 2005) as our seismic texture change indicator, as we are looking for changes to higher frequencies that usually results in chaotic textures.
The attribute we have chosen to implement is a 3-D Sobel based edge detector, namely amplitude contrast (Aqrawi & Boe, 2011). In essence, the textural analysis in this case will decide the outcome of two things. One, the choice of preconditioning prior to filtering, and the second is the normalization method used. While, when implementing the operator size of our edge filter, relative depth/time is used to adjust it adaptively.
A heavily faulted seismic dataset from the Norwegian North Sea has been used to test the methodology of the adaptive calculations. Our results indicate that the adaptive edge method ensures a higher level of detail, and highlights the smaller amplitude discontinuities better than a static operator size approach. It also proves to increase continuity and reduces the detection of noise, which overall gives a more accurate edge detection of the seismic data.
AAPG Search and Discovery Article #90163©2013AAPG 2013 Annual Convention and Exhibition, Pittsburgh, Pennsylvania, May 19-22, 2013