Signal attenuation happens when the amplitudes get reduced in the data, as a result of transmission losses, scattering, inelastic processes and spherical spreading. This is usually compensated for by applying some kind of "gain" (balance, scaling, AGC etc.) overall and by the application of "Q compensation" to deal with the situation where the signal decay is frequency dependant. (High frequencies get scattered/attenuated more than low ones, in general)
The CREWES research group have published some work that suggests the use of Gabor Deconvolution as an alternative to Q compensation, so it may be worth testing these and comparing them. If you have good estimates of Q (from wells) then Q compensation may be an easier option.