Shape Compression using Spherical Geometry Images

Abstract

We recently introduced an algorithm for spherical parametrization and remesh- ing, which allows resampling of a genus-zero surface onto a regular 2D grid, a spherical geometry image. These geometry images offer several advantages for shape compression. First, simple extension rules extend the square im- age domain to cover the infinite plane, thereby providing a globally smooth surface parametrization. The 2D grid structure permits use of ordinary im- age wavelets, including higher-order wavelets with polynomial precision. The coarsest wavelets span the entire surface and thus encode the lowest frequen- cies of the shape. Finally, the compression and decompression algorithms op- erate on ordinary 2D arrays, and are thus ideally suited for hardware ac- celeration. In this paper, we detail two wavelet-based approaches for shape compression using spherical geometry images, and provide comparisons with previous compression schemes.