Reconstruction and simplification of high-quality multiple-region models from planar sections

This paper proposes an accurate and efficient process that reconstructs, smoothes, and simplifies large-scale, three-dimensional models with multiple regions from serial sections. Models are reconstructed using a generally classed marching tetrahedra method that is less complex than the recent generally classed marching cubes algorithms, yet still preserves interface conformability between the regions in the model. Surfaces are smoothed using a volume preserving Laplacian filter that also preserves the region interfaces and topologies. Models are simplified using an efficient and accurate quadric-based edge contraction scheme that maintains the interfaces between regions and preserves the topology of the model. The edge contraction process is constrained to produce surface meshes that have high-quality facet shapes. The process both reconstructs as well as simplifies these models on-the-fly in one pass so that huge models may be processed within limited computer memory. The process does not require the entire original model to fit in the memory at one time. Example results of multiple-region models from the fields of materials science and medicine are presented.