We propose a new hierarchical representation of discrete data sets living on graphs. The approach takes advantage of recent works on graph regularization. The role of the merging criterion that is common to hierarchical representations is greatly reduced due to the regularization step. The regularization is performed recursively with a decreasing fidelity parameter. This yields a robust representation of data sets. We show experiments on digital images and image databases. © 2011 Springer-Verlag.
CITATION STYLE
Hidane, M., Lézoray, O., & Elmoataz, A. (2011). Hierarchical representation of discrete data on graphs. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 6854 LNCS, pp. 186–193). https://doi.org/10.1007/978-3-642-23672-3_23
Mendeley helps you to discover research relevant for your work.