Detection of the tear meniscus shape using asymmetric graph-cuts

Tamir Yedidya, Richard I. Hartley, Jean-Pierre Guillon, Yogesan Kanagasingam

Abstract

We present a new fully automatic algorithm to evaluate the shape and regularity of the tear meniscus in eye images taken using a slit-lamp after instilling fluorescein. Our method analyzes the meniscus in the corneal and conjunctival areas and detects abnormalities such as conjunctival folds. We use graph-cuts to minimize a cost function to simultaneously produce a segmentation of the meniscus and the best shape prior for the eyelids. The pairwise term is asymmetric in order to capture the global properties of the meniscus and add a sense of direction. We tested our method on 43 images and provide a grading of the quality of the meniscus.
BibTeX
@inproceedings{yedidya2010detection,
  author = {Tamir Yedidya and Richard I. Hartley and Jean-Pierre Guillon and Yogesan Kanagasingam},
  title = {{Detection of the tear meniscus shape using asymmetric graph-cuts}},
  booktitle = {ISBI},
  year = {2010},
  doi = {10.1109/isbi.2010.5490143},
}

Loading related content...

Figure