We have used a dense wide-angle data set to test a two-step procedure for the separate inversion of first-arrival and reflection traveltimes. Data were collected in a complex thrust belt environment (southern Italy) along a 14-km line, with closely spaced sources (60 m) and receivers (90 m). We have applied a fully non-linear tomographic technique, specially designed to image complex structures, to over 6400 first-arrival traveltimes in order to determine a detailed velocity model. A bi-cubic spline velocity model parametrization is used. The inversion strategy follows a multiscale approach, and employs a non-linear velocity optimization scheme. The tomographic velocity model is adopted as the background reference medium for a subsequent interface inversion aimed at imaging a target upper-crust reflector. The interface inversion method is also based on a multiscale approach and uses a non-linear technique for model parameters (interface position nodes) optimization. We have applied the interface inversion method to over 1600 reflection traveltimes of a target event picked both in the near- and in the wide-angle offset range. The retrieved interface iswell resolved in the central part of the model, where ray coverage mainly includes clear post-critical reflections and the background velocity model is accurate in depth thanks to large offset deep turning rays. The velocity and interface models thus determined are consistent with Vertical Seismic Profiling data and correlate well with the geometry of known geological structures. This study shows that the used inversion approach is efficient for target-orientated investigations in complex geological environments.