Methods for adaptively separating diffractions

The ranks or the singular (characteristic) values of the matrix data are calculated by the SVD method and are measures of the different dipping events within the matrix data (seismic data). This approach is well-known as the matrix pencil method (MPM) (Jain, 1974; Sarkar and Pereira, 1995; Hua and Sarkar, 1991). The MPM approach has been used for dispersion estimation of sonic logging data (Ekstrom, 1996; Lang et al., 1987), and assumes that the first a few characteristic values are associated with the reflection data while the rest are related to diffractions. Therefore, we can estimate or separate reflections and diffractions from recorded seismic data by matrix rank reduction or truncation. As the seismic events are not necessarily linear or follow plane wave assumption, we propose to apply the matrix rank reduction approach on windowed data to better approximate the plane-wave assumption for the reflection data. We call this method a localized rank-reduction method.

Subsections


2020-12-05