Conventional velocity analysis can handle a horizontally stratified me
dium well. There is no indication, though, that it will be as successf
ul when applied to a more complicated geological structure. In fact, a
small angle of incidence may transform to a wide-angle reflection eve
nt for a dipping interface. In this case, conventional velocity analys
is may lead to large errors and thus cannot be applied. Seismic tomogr
aphy is attractive as it is virtually free from any restrictions impos
ed on the velocity distribution in the model space or on the setup of
a seismic experiment. It is important, however, to recall that seismic
tomograpby yields results of inferior quality compared to medical tom
ography. This paper investigates the reason for this and how to suppre
ss a significant blurring of seismic tomograms. Unlike medical tomogra
phy, one cannot provide full angular coverage of the model space in a
typical seismic experiment: the sources and the receivers cannot surro
und an unknown object inside the earth to provide a complete spectrum
of view angles. Incomplete angular coverage may lead to the occurrence
of large inaccuracies in the computed tomograms especially when the i
nitial model is poorly chosen. We demonstrate a method of suppressing
the adverse effects related to an incomplete angular recording. This i
s 'compensation tomography' which can be used efficiently in the case
of a limited angular aperture. Numerical experiments illustrate the th
eory.