Skip to content (access key 's')
Logo of Technion
Logo of CS Department


Pixel Club Seminar: 3D Shape Representation Using Diffusion Embeddings
event speaker icon
Radu Horaud (INRIA Rhone-Alpes, Grenoble, France)
event date icon
Monday, 31.5.2010, 11:30
event location icon
EE Meyer Building 1061
We will address the problem of representing shapes using eigenvalues and eigenvectors of the discrete diffusion operator. A discrete shape, such as a mesh or a point cloud, can be viewed as an undirected weighted graph, hence one can use spectral graph theory to both embed and analyse shapes. We propose two graph diffusion operators that are built based on two widely used graph Laplacians: The combinatorial Laplacian and the normalized Laplacian. We recapitulate the basic spectral properties of these matrices and we thoroughly motivate the choice of the combinatorial Laplacian. The embedded shape is therefore treated as a distribution in the corresponding embedded (or feature) space. We propose several possible normalizations and we characterize some of their statistical properties. We will illustrate the usefulness of this shape representations for the task of sparse shape matching.
[Back to the index of events]