Compression of textured surfaces represented as surfel sets

Abstract

A method for lossy compression of genus-0 surfaces is presented. Geometry, texture and other surface attributes are incorporated in a uni ed manner. The input surfaces are represented by surfels (surface elements), i.e., by a set of disks with attributes. Each surfel, with its attribute vector, is optimally mapped onto a sphere in the sense of geodesic distance preservation. The resulting spherical vector-valued function is resampled. Its components are decorrelated by the Karhunen-Lo eve transform, represented by spherical wavelets and encoded using the zerotree algorithm. Methods for geodesic distance computation on surfel-based surfaces are considered. A novel e cient approach to dense surface attening/mapping, using rectangular distance matrices, is employed. The distance between each surfel and a set of key-surfels is optimally preserved, leading to greatly improved resolution and eliminating the need for interpolation, that complicates and slows down existing surface unfolding methods. Experimental surfel-based surface compression results demonstrate successful compression at very low bit rates.