Pitfalls of using conventional and discrete Radon transforms on poorly sampled data

Abstract

The least‐square discrete Radon transform (DRT) is currently one of the most popular methods used in the suppression of multiples and other coherent noise events on irregularly sampled data gathers used for prestack true amplitude analysis. Unfortunately, in the absence of a priori information, this technique suffers from the same aliasing problems as Fourier and conventional (τ, p) methods. Although the DRT is able to reconstruct the original image more accurately than conventional (τ, p) transforms, a harmful by‐product is an increase in the amplitude of aliased events in the transform domain. In particular, the DRT will boost the amplitude of the aliases of true events that fall outside the p analysis window to help reconstruct the input data. These amplified aliases degrade signal periodicity in the (τ, p) domain. If muted, they can destroy subtle amplitude changes necessary for amplitude variation with offset (AVO) analysis. At the very least, one should carefully evaluate the choice of analysis window and mutes when designing a filter in the (τ, p) domain. Alternatively, one can exploit additional a priori information based on semblance. Iterative application of the DRT and mutes can also be used to suppress aliased events further.