Figure 4.
The proposed algorithm to extract the centerline from the model. From the input model (1), we compute two types of distances: (2) the distance to the wall (DW) and (3) the distance to the in/outlet (DL). The model is then subdivided (4) by splitting the range of DL. A raw centerline (5) is generated by choosing a node for each piece which has the maximal DW. The raw centerline is divided (6) into segments using node valences. The final centerline is obtained by re-sampling (7) and then super-sampling (8) to improve nodal distribution.