Talk:Graph cuts in computer vision

From Wikipedia, the free encyclopedia

To-do list for Graph cuts in computer vision:

Here are some tasks you can do:
  • Copyedit: Copyedit existing sections
  • Expand:
    • Add a (mostly-)worked example for a specific small binary problem (e.g. denoising a binary image, or foreground/background separation)
    • Add section explaining techniques such as expansion moves to solve k > 2-class problems
    • Maybe a section better explaining the connection to energy minimization?
This article is within the scope of WikiProject Computer science, which aims to create a comprehensive computer science reference for Wikipedia. Visit the project page for more information and to join in on related discussions.
Stub rated as stub-Class on the assessment scale
Low rated as low-importance on the assessment scale


One of my gripes about the article is that it stops right before it gets to one of the best parts of graph cuts: that through constructions like expansion moves (see Boykov, Veskler, Zabih as referenced in the article), you can minimize the energy of a k > 2-class problem. Iknowyourider (t c) 02:48, 22 June 2007 (UTC)

[edit] can someone add a small example?

Can someone give an example of the algorithm working on a small image? a) show how the source, sink and pixels are connected as a graph. b) show how the energy values are computed for one edge. c) show how the system arrives at one cut. —The preceding unsigned comment was added by Hmulling (talkcontribs) 07:20, 25 Jul 2007 (UTC)

Here's some lecture slides that briefly introduce Network flow and Max flow min cut, and then show the reduction of the binary labeling problem to an instance of finding a min cut. The next set of slides have some further details, and a lot of cool example images where the stuff is applied to solve all sorts of problems. The lectures after that [1] [2] introduce k > 2-class problems. I will try to add a small worked example to this article. Feel free to bug me if I haven't done anything in a few days. Cheers, Iknowyourider (t c) 08:11, 25 July 2007 (UTC)