Structure Oriented Signal Enhancement (SOSE)

Abstract

Most existing signal enhancement and random noise removal algorithms take no account of structure, and so tend to smear coherent events across faults and harm steeply dipping events. A new noise suppression technique is presented here which uses 3D structure tensors, calculated from stacks or offset panels, to detect event edges and applies anisotropic diffusion filtering between those edges, to enhance signal at the expense of noise. Using both synthetic and real data, we have demonstrated the capability of this technique to remove noise as effectively as existing methods whilst also preserving signal and honouring truncations (faults, etc). We propose how this fast and flexible technique can be used not only on seismic reflection data as an aid to interpretation, speeding up automated horizon picking, but also on other attribute volumes, for example as a robust method of helping build initial velocity models that honour structure.