Efficient extraction of isosurface components using GPU acceleration and cell decomposion

초록

In this paper, we describe an accurate and efficient algorithm for extracting a contour, a connected component of an isosurface, from piece- wise trilinear volume data defined on rectilinear grids. Contour Tree has been a useful tool for interactive contour selection and extraction. However, conventional contour extraction algorithms require volume data to be defined on simplicial grids, which causes inefficiency in terms of processing time and output mesh size. Our contour extraction algorithm provides an accurate and efficient way to work from the rectilinear volume data without changing the topology of trilinear function by performing topology preserving cell subdivision only when necessary. We achieved substantial improvements to the extraction performance by minimizing expensive cell subdivision and utilizing many-core GPU architectures. © 2015 International Information Institute.

키워드