Scale-invariant OpenCV image template matching in Python

Asked Nov 15 '21 at 16:06

Active Nov 15 '21 at 16:06

Viewed 2,596 times

The method cv2.matchTemplate gives a translation-invariant template matching on an image, but not scale-invariant.

How to do a scale-invariant (or even rotation-invariant) template matching with Open CV for Python?

Is there a way to use a log-polar FFT template matcher built-in in OpenCV for Python?

(here is a C++ version)

Note:

I'd like to avoid the "simple trick" of matching many times in loop with scaled versions of the same template, as shown in Multi-scale Template Matching using Python and OpenCV
Related: https://answers.opencv.org/question/169613/how-does-macthtemplate-deal-with-scali
https://pypi.org/project/imreg/ is related but no longer developed and with no documentation

asked Nov 15 '21 at 16:06

Basj

1

See https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.730.9105&rep=rep1&type=pdf and https://scikit-image.org/docs/stable/auto_examples/registration/plot_register_rotation.html – fmw42 Nov 15 '21 at 16:34
@fmw42 Thank you for the links. For the second, do you have a code example about how it can be applied to a template matching context? For the first, is there a code sample repo somewhere? – Basj Nov 15 '21 at 16:48
feature matching is popular... – Christoph Rackwitz Nov 15 '21 at 18:09
@Basj. Sorry, no other examples for either link. I tried coding to extend the second and got one example to work, but it failed on another. – fmw42 Nov 15 '21 at 18:31

0 Answers0