Manifold and Data Filtering on Graphs
Résumé
High-dimensional feature spaces are often corrupted by noise. This is problematic for the processing of manifold and data since most of reference methods are sensitive to noise. This paper presents preprocessing methods for manifold denoising and simplification based on discrete analogues of continuous regularization and mathematical morphology. Both approaches enable to project the data onto a submanifold with a graph generated by the data.
Domaines
Traitement des images [eess.IV]
Origine : Fichiers produits par l'(les) auteur(s)