[en] Image reconstruction of 3D cone-beam tomography using the FDK algorithm requires huge computational resource. The reconstruction of a high resolution 3D object would usually take a long time, which is intolerable in real applications. Parallel computing using cluster systems is one of the common approaches to solve this problem. The mixed model with both MPI and Pthreads is used to parallelize the FDK algorithm on a SMP cluster. The computation is parallelized in a two-level manner with both Message-passing among nodes and Shared-Memory within each node. Our simulation results on the SMP cluster system are reported. (authors)