0

I am looking for some guidance here. I am trying to compare two x-rays of jaw bones. The x-rays show trabecula bone, which is bone that contains lots of holes. I am hoping to show that trabecular bone patterns, like fingerprints, can be used to identify an individual. I want to use the Hausdorff distance to compare how similar (or different) two x-rays are. I take the x-rays and threshold them to become black/white images, then use an edge detection filter. I wrote a C# program to compute the Hausdorff distance of the two images. Can I identify the points in both images as starting at the origin (0,0) or do they need to be spaced at some distance from each other? Thanks in advance for your help. Jim

user2183837
  • 1
  • 3
    I would expect that you would first need to deal with translation, rotation, and scaling discrepancies since otherwise the two images my have the same feature at slightly different coordinates. Random paper for doing all three at once using a log-polar transform.: http://www-cs.engr.ccny.cuny.edu/~wolberg/pub/icip00.pdf – Eric Towers Sep 16 '20 at 18:32
  • To all the transformations listed by Eric Towers that are called "registration", you can add hopefully a conical (or perspective) effect. Besides, you have to take into account the histogram of gray levels in order to adapt the threshold to each image. Only after that you can consider Hausdorff distance. – Jean Marie Sep 16 '20 at 20:35
  • I am not sure at all Hausdorff distance is interesting in this case. It will give you a distance but how do you want to use it ? Hausdorff distance between the two shapes will have to be done on small parts of your image, for at least two reasons (being meaningful, having a not to long computation time)... – Jean Marie Sep 16 '20 at 20:43

0 Answers0