Reliable 3-D velocity estimation is a prerequisite to 3-D migration in laterally varying environments, but is difficult with data from conventional seismic surveys because of limited aperture, especially in the crossline direction. A possible solution to this problem is tomographic estimation of the 3-D velocity distribution using transmitted (turned) rays recorded in wide areal apertures. Using this approach, the shape and position of salt structures, including overhangs, sub salt features, and deformation of surrounding sediments, are accurately estimated from synthetic test data. Model enhancement based on a priori geologic constraints gives 3-D macro velocity distributions for models of three actual salt structures: a salt tongue in the Gulf Coast, and the Hainesville and Mount Sylvan domes in the East Texas Basin. The geometries required are feasible with current land acquisition systems. The main practical limitations are the requirements for nonstandard (wide aperture) recording arrays and for accurate time picking of the transmitted arrivals.