Additional to what Joachim Pimiskern already said, I would assume that an IEC-based (interactive evolutionary computing) approach, based on interactive fitness (subjective user assessment) and different populations, might be feasible - depending on your overall processing goal. I got this idea in mind for quite some time now: IEC for interactive image pre-processing and design of domain-dependent basic feature-extraction-chains ... still not 100% sure, if this is a promising approach ...
@Joachim Pimiskern has made a very good suggestion, since ants naturally will want to follow some path and the edges in an image provide lots of choices for paths for ants to follow. In fact, we can Joachim's idea a step further, and prompt artificial ants to look for edges along the border of potential convex sets or convex hulls (as "envelopes of image pixels) or convex cones (for images containing corners).
Perhaps the comment "how to detect an edge" by Sandeep can be rephrased to read "how to detect interesting edges", since the ant following paradigm works best if we edge detection has been before we invoke an ant following algorithm. Do you agree, Sandeep or Joachim?
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing
- Image Processing