Image edge enhancement and segmentation via randomized shortest paths

This paper describes a new method for image edge enhancement and boundary segmentation. Like many interactive graph-based segmentation methods, users are asked to provide some foreground (or object) and background seeds. A set of randomly generated points representing the foreground are paired with...

Full description

Saved in:
Bibliographic Details
Published in2012 5th International Conference on Biomedical Engineering and Informatics pp. 290 - 294
Main Authors Ming Xu, Jun Wang, Zeyun Yu
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.10.2012
Subjects
edge enhancement
image segmentation
Monte Carlo method
shortest paths
Online AccessGet full text

Cover

More Information
Summary:This paper describes a new method for image edge enhancement and boundary segmentation. Like many interactive graph-based segmentation methods, users are asked to provide some foreground (or object) and background seeds. A set of randomly generated points representing the foreground are paired with another set of random points representing the background. The corresponding shortest paths of all such pairs are found and accumulated. These paths tend to go through the boundaries of the object of interest. Therefore, the accumulated paths can enhance the object edges, from which the final segmentation is obtained. Several experiments are provided to demonstrate the effectiveness of the proposed approach.
ISBN:9781467311830
1467311839
DOI:10.1109/BMEI.2012.6512925