Spherical parameterization and geometry image-based 3D shape similarity estimation (CGS 2004 special issue)

Abstract

In this paper, we describe our preliminary findings in applying the spherical parameterization and geometry images to the task of 3D shape matching. View-based techniques compare 3D objects by comparing their 2D projections. However, it is not trivial to choose the number of views and their settings. Geometry images overcome these limitations by mapping the entire object onto a spherical or planar domain. We make use of this property to derive a rotation invariant shape descriptor. Once the geometry image encoding the object’s geometric properties is computed, a 1D rotation invariant descriptor is extracted using the spherical harmonic analysis. The parameterization process guarantees the scale invariance, while its coarse-to-fine nature allows the comparison of objects at different scales. We demonstrate and discuss the efficiency of our approach on a collection of 120 three-dimensional models.