Unifying framework for 2D/3D shape classification based on the eigenvectors of the Laplace–Beltrami Operator and wavelet density estimation.
Form prototype representations of shapes on a high dimensional spherical manifold using the Karcher mean.
Computationally efficient classification scheme due to smaller search index composed of mean shape densities.
Spherical manifold geometry has a simple closed form similarity metric.
Performs well on several benchmark datasets.