In order to extract faults from seismic data, an attribute or a series of attributes must reflect at least one seismic expression of a fault. The expression of a fault can vary significantly from amplitude variations to subtle discontinuities, flexures and sharp discontinuities. Identification of these expressions through varying edge detection attributes results in a better fault extraction.
Structurally Oriented Semblance is the fastest of the three algorithms and is very good at picking up various types of faults within the data: large scale versus small scale. Faults with a different seismic expression either side of the fault can be detected with this algorithm.
The Tensor attribute is based on a local structural tensor which is analysed to find the dominant direction of the reflectors and their structure. Tensor images faults that are expressed with Flexure particularly well as well as small scale faults.