Efficient graph cuts for unsupervised image segmentation using probabilistic sampling and SVD-based approximation


Keuchel, Jens ; Schnörr, Christoph


[img]
Preview
PDF
2003_9.pdf - Published

Download (1MB)

URL: http://ub-madoc.bib.uni-mannheim.de/1805
URN: urn:nbn:de:bsz:180-madoc-18051
Document Type: Working paper
Year of publication: 2003
Publication language: English
Institution: School of Business Informatics and Mathematics > Sonstige - Fakultät für Mathematik und Informatik
MADOC publication series: Veröffentlichungen der Fakultät für Mathematik und Informatik > Institut für Informatik > Technical Reports
Subject: 004 Computer science, internet
Subject headings (SWD): Bildverarbeitung , Nyström-Verfahren , Singulärwertzerlegung
Abstract: The application of graph theoretic methods to unsupervised image partitioning has been a very active field of research recently. For weighted graphs encoding the (dis)similarity structure of locally extracted image features, unsupervised segmentations of images into coherent structures can be computed in terms of extremal cuts of the underlying graphs. In this context, we focus on the normalized cut criterion and a related recent convex approach based on semidefinite programming. As both methods soon become computationally demanding with increasing graph size, an important question is how the computations can be accelerated. To this end, we study an SVD approximation method in this paper which has been introduced in a different clustering context. We apply this method, which is based on probabilistic sampling, to both segmentation approaches and compare it with the Nyström extension suggested for the normalized cut. Numerical results confirm that by means of the sampling-based SVD approximation technique, reliable segmentations can be computed with a fraction (less than 5%) of the original computational cost.
Additional information:

Das Dokument wird vom Publikationsserver der Universitätsbibliothek Mannheim bereitgestellt.




+ Citation Example and Export

Keuchel, Jens ; Schnörr, Christoph (2003) Efficient graph cuts for unsupervised image segmentation using probabilistic sampling and SVD-based approximation. Open Access [Working paper]
[img]
Preview


+ Search Authors in

+ Download Statistics

Downloads per month over past year

View more statistics



You have found an error? Please let us know about your desired correction here: E-Mail


Actions (login required)

Show item Show item