Some Theoretical Links Between Shortest Path Filters and Minimum Spanning Tree Filters

Abstract : Edge-aware filtering is an important pre-processing step in many computer vision applications. In literature, there exist several versions of collaborative edge-aware filters based on spanning trees and shortest path heuristics which work well in practice. For instance, Tree Filter (TF) which is recently proposed based on a minimum spanning tree (MST) heuristic yields promising results in many filtering applications. However, links between the tree-based filters and shortest path-based filters are faintly explored. In this article, we introduce an edge-aware generalization of the TF, termed as UMST filter based on the union of all MSTs. The major contribution of this paper is establishing theoretical links between filters based on MSTs and filters based on geodesics via power watershed framework. More precisely, we show that Union of Minimum Spanning Trees (UMST) filter can be obtained as the limit of Shortest Path Filters (SPFs). Intuitively, TF can be viewed as an approximate limit of the SPFs. We propose and provide a detailed analysis of two different implementations of the UMST filter based on shortest paths. Further, we establish empirically with the help of denoising experiments that TF is an approximate limit by showing that TF and one of our approximations yield similar results.
Type de document :
Article dans une revue
Journal of Mathematical Imaging and Vision, Springer Verlag, In press
Liste complète des métadonnées

https://hal.archives-ouvertes.fr/hal-01617799
Contributeur : Sravan Danda <>
Soumis le : jeudi 22 novembre 2018 - 06:31:41
Dernière modification le : jeudi 6 décembre 2018 - 01:15:31

Fichier

PTF_JMIV_2.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

  • HAL Id : hal-01617799, version 6

Citation

Sravan Danda, Aditya Challa, B Daya Sagar, Laurent Najman. Some Theoretical Links Between Shortest Path Filters and Minimum Spanning Tree Filters. Journal of Mathematical Imaging and Vision, Springer Verlag, In press. 〈hal-01617799v6〉

Partager

Métriques

Consultations de la notice

22

Téléchargements de fichiers

13