Layered Depth-Normal Images for complex geometries - Part two: Manifold-preserved adaptive contouring

We present an adaptive contouring approach to generate contour surface from solid models represented by Layered Depth-Normal Images (LDNI) sampled in three orthogonal directions. Our contouring algorithm builds an octree structure for mesh generation in a top-down manner: starting from the bounding box of a LDNI solid model, the cells are recursively subdivided into smaller sub-cells based on the topology and geometry criteria of refinement until both of the requirements, the topology in cell is simple and the geometry approximation error is less than a user defined tolerance, are satisfied. The subdivision also stops when the processed cells reach the finest resolution of LDNI models. In order to overcome the topology ambiguity inside a cell that leads to the occurrence of nonmanifold entities, we analyze the possible inside/outside configurations of cell-nodes and exploit two strategies to generate manifold-preserved mesh surfaces. Moreover, the most time-consuming step of our contouring algorithm - the construction of octree structure can be easily parallelized to run under a computer framework with multiple-processors and shared memory. Several examples have been tested in the paper to demonstrate the success of our method.