Signal/Noise Separation with Slant Stacks and Migration
, by William Harlan
Geologic events strongly color a seismic section spatially. Removable noise is spatially white and non-gaussian (parsimonious). Transformations, such as slant stacks and migration, exist which make geology parsimonious. To estimate a parsimonious signal from a gaussian background, one should zero low amplitudes, which are not parsimonious. One should estimate geology after transformation and noise before--removing one from the data before estimating the other. Estimations with a slant stack separate real data into geology and noise very well. Similar existing algorithms (e.g. using slant stacks to prevent aliasing) do not make local estimations, thereby failing to remove noise or introducing artifacts.
STANFORD