Noise Expression Filters
Noise Expression utilises ffA’s example driven parameter optimisation approach, which enables you to obtain an optimised result very quickly and efficiently. Three different workflows are available: Random Noise, Coherent Noise and Aggressive Noise.
The Random Noise workflow uses the TDiffusion filter, which is a noise reducing, adaptive edge preserving filter targeting random noise. The idea is to repeatedly process a volume with a low pass filter to produce a series of increasingly smoother, images by solving a Partial Differential Equation through time. The filter is anisotropic since it adapts to local orientation and adapts the direction of the filter. The TDiffusion filter differs from the SO filters described previously in that the orientation information embedded in the structural tensor enables the TDiffusion.
The Coherent Noise workflow uses the SO FMH (Structurally Oriented: Finite Median Hybrid) filter, which is a noise reducing, edge-preserving filter. It removes coherent noise, such as, minor acquisition or migration noise whilst preserving subtle details like edges, corners and sharp dips in the structure. Minor acquisition or migration noise Structurally Oriented Noise filtering retains much more detail than traditional smoothing techniques, as it uses Dip and Azi volumes to steer the smoothing filter.
The Aggressive Noise workflow uses the SO Noise filter, which is a Structurally Oriented operation and targets aggressive noise such as salt or basalt effects. It is not as sophisticated as the SO FMH or TDiffusion filter and may not maintain edges. It retains much more detail than traditional smoothing techniques as it does not smooth across the geology. The Structurally Oriented noise filters can be applied along an orientation defined by the local structural vector at each point in the data.
Many data sets require attenuation of both coherent and random noise; in these instances it is imperative to run the SO FMH filter before the TDiffusion filter. The TDiffusion filter will ‘see’ any coherent event as signal, thus preserving noise within the data. If the coherent noise is attenuated with the SO FMH filter the TDiffusion filter will attenuate the remaining random noise. For the Noise Expression tool simply generate the SO FMH first and use that as the input back into the Noise Expression tool to calculate the TDiffusion.