I use SIFT, and I want to find the angle of inclination of an image compared to two reference images, in short:

I have 3 images :

Image x: input image

Two images 1 and 2, reference images that are very similar.

My goal is to find the angle of rotation between the image x and a combination between the two images 1 and 2. My problem is that I dont know how to combine the two images of the database to compare them with to the image x.

The purpose of the combination is to extract a vector Keypoints and Mat descriptor from the 2 keypoint vectors and the 2 Mat descriptors.

enter image description here

  • $\begingroup$ The obvious answer is to simply add them taking care of saturation. But this might complicate things depending on the nature of the two images. Is it possible to provide some more information about these two images? How about running SIFT twice and then getting the average between the two point estimates (?). $\endgroup$
    – A_A
    Oct 31 '19 at 10:41
  • $\begingroup$ Sorry my diagram was not very clear i edit it now $\endgroup$
    – all.west
    Oct 31 '19 at 11:02
  • $\begingroup$ The images are an image of a valve and there is a QR code next the valve $\endgroup$
    – all.west
    Oct 31 '19 at 11:03
  • $\begingroup$ ...I think that your previous diagram was fine. First combine the images, then run a single SIFT on them. The QR code's "features" are so stark they will simply persist. Except of course if you accidentally super impose them on top of some wall or ground texture that is like crumbled alimunium foil, that would throw the whole thing off pretty quickly. $\endgroup$
    – A_A
    Oct 31 '19 at 11:05
  • 1
    $\begingroup$ That is odd...So do you expect that you will match up the two images, then subtract them....and get mostly the background? The reason I am asking what are you trying to do is because it might be possible to look at the problem differently and come up with a simpler or more manageable solution. $\endgroup$
    – A_A
    Oct 31 '19 at 14:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.