Fast and accurate 3D terrain reconstruction to acquire a high resolution Digital Elevation Model (DEM) is one of the most important research areas in geographic information representation. As a kind of spatial interpolation method that is superior to other accurate interpolation methods, the Radial Basis Function (RBF) is particularly suitable for the reconstruction of complex 3D terrain models. However, the efficiency when calculating this interpolation model becomes lower and lower with an increasing number of sampling points, and the interpolation equation becomes too difficult or even fails as the interpolation matrix become bigger and bigger. To address this issue, a parallel interpolation method based on the principle of domain decomposition and restricted additive schwarz method (referred to as RASM) is proposed. A compact support RBF (CSRBF) based global interpolation matrix was built by taking all of the known sampling points, and the optimal local compact support radius is calculated for each of the local domains. The interpolation procedure operates in parallel through a message passing interface (MPI) based on the RASM. DEM data are used in an interpolation experiment. The results show that the method proposed in this paper could accurately reconstruct the terrain with massive terrain sampling data enabling a high efficiency solution.