Well, that's what it says in the final report.
"we applied an empirical process to extrapolate reflectivity coefficients in the frequency-wave number domain with a view to preserving amplitude relationships and finalised the correlated data set with a horizon dependent, cross-line, predictive deconvolution and subsequent FX-filtering to reduce multiples, enhance continuity and emphasise micro-faulting in the target area"
(which is absolute rubbish)
Funny - now I'm working predominantly for the marketing and advertising industry, I hardly notice the difference.)