Conventional algorithms for tomographic reconstruction require the acquisition of a complete set of projections at uniform angular displacements. In many cases, however, the geometry of the sample or a loss of data can significantly reduce the range of the available projections. Several algorithms have been proposed in literature to handle such situations, but their performances are low or they require strong constraints and hypothesis about the nature of the sample or the data. Here a new method is proposed. It is based on a novel morphing technique, which affords in general terms the problem of curve matching and is here specialized to the case of tomographic reconstruction. The proposed algorithm is very fast in comparison to other approaches having similar effectiveness; furthermore, it allows one to obtain good quality images even when a significant fraction of the views is absent, without any hypothesis about the nature of the sample or the kind of measurement. The results obtained by applying this technique to the Shepp-Logan phantom and to a clinical scan are reported here and discussed.