Multi-views reconstruction based on cell primitives

In this paper a new algorithm is presented for constructing solid model from multiple projections. It is a combination of wireframe-oriented and volume-oriented approach. The key steps are the partition and generation of cell primitives. The general translational sweeping operation is used to restore topologic and geometric information simultaneously, and overcome some deficiencies of previous algorithms. Furthermore, when there are multiple interpretations of input views, users are allowed to involve in the process of reconstruction to select or confirm the interpretation. They could interact in reconstruction process in different granularity levels such as primitive level, baseface level. The practicability of algorithm is effectively improved by this approach.