Improvement of the signal to noise ratio of seismic traces by re-alignment of reverberant energy

Marine seismic sections recorded in areas having a hard sea-bottom are characterised by high levels of reverberant energy, as in many North Sea sections for example. Reverberant energy is undesirable from a seismic interpretation viewpoint since it interferes with the identification of the deeper substructure. Conventional processing attempts to remove or attenuate the reverberant energy. Conceptually at least, this is wasteful since each reverberant component (multiple) contains coherent signal energy which, if suitably re-aligned, could be used to enhance the trace signal to noise ratio (SNR). Any such re-alignment is complicated by the fact that successive components are, in general, overlapping and distorted, and the interval between successive multiples is only approximately known. Nevertheless, the aim of this paper is to show that it is feasible to extract the coherent energy from multiple components in such a way as to make possible improvements in SNR by averaging the extracted components with the original. We further demonstrate that the SNR enhancing function may be combined with conventional predictive filtering.