“…Since the entire shape is used for retrieval, normalization techniques can be used to remove much of the transformational ambiguity in matching, allowing for the use of the center of mass for removing translational ambiguity, radial-variance or mean-/bounding-radius for removing scaling ambiguity, and principal axes for rotational ambiguity. These methods have included: 1D histograms capturing the distribution of points [2,3,4], crease angles [5], and curvature [6] over the surface; spherical functions characterizing the distribution of surface normals [7], axes of reflective symmetry [8], conformality [9], and angular extent [10]; 3D functions characterizing the rasterization of the boundary boundary points [11] and the distance transform [12]; and even 4D plenoptic functions characterizing the 2D views of a surface [13].…”