The need for increased resolution in velocity spectra is clear when one wishes to distinguish between neighboring primary events from reflectors with conflicting dip, or to identify primaries in the presence of multiples. The transformation in velocity analysis from the offset and reflection-time domain to the stacking velocity and zero-offset-time domain can be achieved using any of several coherence measures based on the crosscorrelations between traces in a collection such as a common-midpoint (CMP) gather or common-image gather (CIG). Use of selected subsets of crosscorrelations, rather than all possible ones in a gather, however, can improve both the reliability and resolution of velocity analysis. In selective-correlation velocity analysis, we include in the summation only those crosscorrelations for whose pair of traces the relative differential moveout of reflections exceeds a chosen threshold value. Comparisons of the performances on synthetic CMP gathers show that selective-correlation velocity analysis considerably enhances the resolving power of velocity spectra over that of conventional crosscorrelation sum (whether normalized or unnormalized) in the presence of closely interfering reflections, statics distortions and random noise, at no sacrifice in the quality of results, and does so at computational cost that is comparable to that for conventional velocity analysis.

No abstract