Zur Hauptnavigation wechseln Zur Suche wechseln Zum Hauptinhalt wechseln

Reeb Graphs Through Local Binary Patterns

Publikation: Beitrag in Buch/Bericht/Konferenzband › Konferenzbeitrag › Begutachtung

Abstract

This paper presents an approach to derive critical points of a shape, the basis of a Reeb graph, using a combination of a medial axis skeleton and features along this skeleton. A Reeb graph captures the topology of a shape. The nodes in the graph represent critical points (positions of change in the topology), while edges represent topological persistence. We present an approach to compute such critical points using Local Binary Patterns. For one pixel the Local Binary Pattern feature vector is derived comparing this pixel to its neighbouring pixels in an environment of a certain radius. We start with an initial segmentation and a medial axis representation. Along this axis critical points are computed using Local Binary Patterns with the radius, defining the neighbouring pixels, set a bit larger than the radius according to the medial axis transformation. Critical points obtained in this way form the node set in a Reeb graph, edges are given through the connectivity of the skeleton. This approach aims at improving the representation of flawed segmented data. In the same way segmentation artefacts, as for example single pixels representing noise, may be corrected based on this analysis.
OriginalspracheEnglisch
TitelGraph-Based Representations in Pattern Recognition
Untertitel10th IAPR-TC-15 International Workshop, GbRPR 2015, Beijing, China, May 13-15, 2015. Proceedings 10
Herausgeber*innenBin Luo, Walter G. Kropatsch, Cheng-Lin Liu, Jian Cheng
ErscheinungsortBerlin Heidelberg
VerlagSpringer International Publishing
Seiten54-63
Seitenumfang10
ISBN (elektronisch)9783319182230
ISBN (Print)9783319182230
DOIs
PublikationsstatusVeröffentlicht - 2015
Extern publiziertJa

Publikationsreihe

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Band9069
ISSN (Print)0302-9743
ISSN (elektronisch)1611-3349

Wissenschaftszweige

  • 102 Informatik

Dieses zitieren