0

I’m looking for a method to do template matching (in Python), where the template is rotated (the angle is unknown). The scale is not modified: I just need to find the rigid body motion. The template is in fact a crop of the original image, rotated by an unknown angle.

I've tried feature matching approaches, using SIFT, SURF, FAST, etc. The problem is that the template is small (around 90x90 pixels) and the keypoint detectors don't find enough strong keypoints (they filter them out). If I relax the constraints and take the keypoints anyway, when I search for point correspondences, I have a lot of outliers. If a filter the outliers with RANSAC, for example, again I have not enough point correspondences to find a transformation.

The problem doesn't seem that difficult, given the template is originated from the original image (crop) and then rotated.

Is there a method I didn't contemplate that is generally used for this kind of problems?

JuaniL
  • 107
  • 1
  • 1
  • 8
  • Please read and follow the posting guidelines in the help documentation, as suggested when you created this account. [On topic](http://stackoverflow.com/help/on-topic) and [how to ask](http://stackoverflow.com/help/how-to-ask) apply here. Since you've already looked at a lot of general solutions, you need something more specific to your problem. We can't supply this without a specific problem presentation. – Prune Apr 10 '18 at 17:01
  • Hi @Prune. With all respect, I think that I described all the details of the problem, and that the question is indeed specific. The answer may be something as specific as an OpenCV function. I've modified the final line to improve the problem presentation. – JuaniL Apr 10 '18 at 18:33

0 Answers0