An efficient framework for constructing multiresolution models from unorganized points

Yong-Jin Liu
DOI: https://doi.org/10.14711/thesis-b651870
1999-01-01
Abstract:Most models reconstructed from unorganized points use dense triangle mesh representations. However, dense triangle-mesh models are awkward to display, transmit, and render at high speed. As a result, we present an efficient framework for easily building a multiresolution model directly from a set of unorganized points sampled from an arbitrary complex object. In our framework, we create output multiresolution models from a class of meshes with subdivision connectivity. Rather than constructing a smooth parameterization for an initial mesh over the base domain, we present a simple ray-intersection approach to obtain a mesh with subdivision connectivity, which we can then easily extend to multiresolution form. Our framework efficiently incorporates various surface-modeling techniques required for creating multiresolution models from sets of unorganized points. Previously, several separate algorithms had to be independently used. As a result, our framework simplifies the complete process of converting sets of unorganized points into multiresolution models and requires substantially less computational time than previously reported algorithms. We have implemented our complete framework on a 450 MHz Pentium II PC with 256 MB of memory. Results show that our integrated framework is efficient and easy to implement.
What problem does this paper attempt to address?