Reducing the number of wavelet coefficients by geometric partitioning

AbstractWith the growing interest toward Internet-based graphic applications, the design of a scalable mesh compression scheme has become a key issue. Using the multi-scale transformation theory introduced by Lounsbery et al. (1997) along with the parameterization techniques of Eck et al. (1995) provides an elegant theoretical framework for producing compact multi-scale representations of surfaces. However, this approach fails to provide good compression and geometric faithfulness in all cases. To solve this problem, we propose a three-step method enabling efficient scalable compression of arbitrary mesh with faithful representations at any level of detail: a partitioning stage along with a triangulation enable the production of a base mesh which preserves the geometry of the model. Then an adaptive parameterization is constructed over this base mesh