ML Anthology

Threshold Selection as a Function of Region Count Stability

Li Yu, Adam W. Hoover
CVPR 2004 pp. 59
doi:10.1109/CVPR.2004.463 /cvpr/2004/yu2004cvpr-threshold/

Abstract

In this paper we present a novel method for threshold selection. The idea is based upon quantifying the stability of the number of regions segmented as the threshold is varied. We capture this idea using a scale-space formulation, and detect "stable" segmentations as local minima in the scale-space. This work was originally motivated by the problem of detecting some types of common lesions in retinal images (many lesions appear as abnormally bright areas), on which we show some results. We also compare our method against an approach based on saliency.

CVPR Semantic Scholar

Cite

Text

Yu and Hoover. "Threshold Selection as a Function of Region Count Stability." IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2004. doi:10.1109/CVPR.2004.463

Markdown

[Yu and Hoover. "Threshold Selection as a Function of Region Count Stability." IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2004.](https://mlanthology.org/cvpr/2004/yu2004cvpr-threshold/) doi:10.1109/CVPR.2004.463

BibTeX

@inproceedings{yu2004cvpr-threshold,
  title     = {{Threshold Selection as a Function of Region Count Stability}},
  author    = {Yu, Li and Hoover, Adam W.},
  booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  year      = {2004},
  pages     = {59},
  doi       = {10.1109/CVPR.2004.463},
  url       = {https://mlanthology.org/cvpr/2004/yu2004cvpr-threshold/}
}