1

I want to compare images to give a sort of score for similarity to a target image, as in the example below (scores just chosen to give an idea of the sort of similarity I'm after.) The scores represent a sort of edit distance. cat image

All the images would be short words or sets of characters like this, and just black and white. The number of elements could vary- e.g. the bottom one has 9 elements but it could be different. The comparison needs to be pretty fast- fractions of a second, which I assume rules out an OCR-based approach.

Is there anything I could use in a library like AForge for something like this (template matching maybe...?) or is it a harder problem? I guess I could look at classifiers but I'd rather not go down that route just yet.

user814425
  • 605
  • 8
  • 19
  • you may also try http://stats.stackexchange.com/ – Leo Jul 31 '14 at 23:52
  • also take a look at http://stackoverflow.com/questions/4196453/simple-and-fast-method-to-compare-images-for-similarity – Leo Jul 31 '14 at 23:56
  • Thanks both, I'm having some promising results from using the Hausdorff distance, as mentioned in this thread http://stackoverflow.com/questions/17129037/distance-between-two-shapes-areas-in-java/23426563#23426563: – user814425 Aug 10 '14 at 15:39

0 Answers0