I am trying to follow this paper [1]. There is a relationship between Markov Random Fields (MRF) to max-flow-min-cut. An MRF can be represented as an undirected graph, and you can find flow through it, but what extra special properties do you get if you know the graph is Markovian? What is the point of putting these two ideas together?

  1. Kwatra, Vivek, et al. "Graphcut textures: image and video synthesis using graph cuts." ACM Transactions on Graphics (ToG). Vol. 22. No. 3. ACM, 2003.
  • 1
    $\begingroup$ Can you quote the passage that suggests this relationship? From skimming title and abstract they seem to use "optimal" cuts to determine where to stitch different images together; if they use max-flow algorithms for that, there does not have to be any relationship to where the graphs come from. $\endgroup$ – Raphael Jul 19 '16 at 14:20
  • $\begingroup$ In the introduction it says the idea of perceptual similarity is close to the idea of MRFs, then the MRF is augmented by special terminal nodes and max-flow/min-cut is found. I was wondering if there's somethings special about the graph being an MRF that makes max-flow/min-cut better, easier to calculate, or if there's some special meaning you can glean from it. But perhaps not. $\endgroup$ – eternalmothra Jul 20 '16 at 14:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.