On Generating Multi-resolution Representations of Polygonal Meshes

Multi-resolution representation is one of the effective approaches to image processing in both computer vision and computer graphics. However, the generation of multi-resolution representation for complex 3D polygonal mesh is a difficult problem. In this paper, a number of tools developed in our research group for this purpose are described. These tools employ different techniques such as smoothing, edge length normalization and wavelets to create representations which describe 3D shape at multi-resolutions. Unlike the control of mesh resolution known as mesh simplification or mesh decimation in computer graphics, the tools we have developed are in the context of computer vision, in particular, object recognition and surface matching. Application examples are given for each of the representations generated by the above tools. Comparison of the representations is also discussed.