Q-guided wavelet-domain amplitude correction

Abstract

This paper presents a novel way of treating seismic reflection amplitudes for transmission effects due to the overburden. The process has benefits for structural interpretation due to the improved visibility of events. Further processing of the data also benefits from a reduction in overburden-induced amplitude changes that can otherwise cause processing artefacts such as migration smiles. A wavelet transform decomposition of the data is used to derive local frequency, space and time varying gain factors from windows of seismic data. These are then constrained using a frequency-constant residual Q model in order to stabilise them against noise or other sources of estimation error. The resulting gain factors are applied to each wavelet voice and the data is reconstructed by an inverse wavelet transform. This is referred to as a Q-guided amplitude correction, as shorthand for the process of constraining the data-derived scalars by frequency-constant residual Q. A by-product of the process is that local estimates of relative Q are generated at the density of the original data volume, and can help with the overall interpretation of the data. Furthermore, once local amplitude variations have been removed, a coarsely sampled estimate of the background Q model can be used to compensate for any slowly varying amplitude and phase changes that exist throughout the survey area.