Spatial dealiasing of 3-D seismic reflection data

Abstract

SUMMARY One of the main challenges of 3-D reflection seismology is providing the spatial sampling required to avoid aliasing. In practice, fine-scale and regular acquisition geometry is possible only in 2-D seismic surveys. 3-D surveys typically have sparse and irregular geometry that often results in spatial aliasing. In this paper we present a processing method to overcome this problem that is suitable for sparse and irregular acquisition geometry, and yields better results than does conventional processing such as dip moveout. The objective of the process is to obtain the hypothetical data that would be recorded in an adequately sampled zero-offset experiment, that best fits the data with a given velocity model. The process takes advantage of what seems to be redundancy in multifold seismic data, to overcome spatial aliasing which is a missing data problem. Our computer implementation involves pre-processing of gain, normal moveout removal, log-stretch transform and Fourier transform. After the log stretching, the relation between the data and the model is time invariant, making inversion practical in the space-frequency (X-F) domain, using a linear inversion, frequency by frequency. After the inversion, the data are inverse-Fourier and inverse-log transformed. The result is ready for post-stack migration.