A Gromov-Hausdorff Framework with Diffusion Geometry for Topologically-Robust Non-rigid Shape Matching

摘要: In this paper, the problem of non-rigid shape recognition is studied from perspective metric geometry. particular, we explore applicability diffusion distances within Gromov-Hausdorff framework. While traditionally used geodesic distance exploits shortest path between points on surface, averages all paths connecting points. The constitutes an intrinsic which robust, in to topological changes. Such changes form shortcuts, holes, and missing data may be a result natural deformations as well acquisition representation noise due inaccurate surface construction. presentation proposed framework complemented with examples demonstrating that addition relatively low complexity involved computation points, its matching performances favorably compare classical presence shapes.