Gaussian beam depth migration of wavelet compressed seismic data

SUMMARY

We have investigated an implementation of Gaussian beam depth migration which images seismic data after being compressed by a digital wavelet transform. Conventional Gaussian beam migration consist of two main steps: (1) a local plane wave decomposition of the input seismic traces into directional components or beam traces; (2) mapping of each beam trace sample to its depth location through dynamic ray tracing into a given velocity model.

Instead of migrating each beam time sample individually, we propose to migrate a selection of picked wavelets. These wavelets are associated with the largest coefficients of an appropriate digital wavelet transform of the beam traces. Coupling a Gaussian beam decomposition with a wavelet compression can lead to an improved signal-to-noise ratio and also a direct saving in computation at the cost of some reduction in imaging accuracy.

We have tested this approach on both synthetic and real data sets. By only imaging a small percentage (1% ÷ 3%) of the wavelet coefficients we have enhanced the signal-tonoise ratio of both stack images and pre-stack common image gathers.