0

I’m currently working on an image registration algorithm which uses aerial imagery. My objective is to compute the percentage of overlap between two images as shown below.

Image 1

Image 2

Image 3

Visually the images have about 50% overlap, I'm using OpenCV following this implementation and this formula, the registered images are warped making it's implementation tricky. A similar implementation with a different formula can be found here.

Are there any simpler workarounds to just find a rough estimate of the overlap?

The Singularity
  • 2,428
  • 3
  • 19
  • 48
  • What’s wrong with the formulas you link to? You don’t cite any output and explain why it’s not what you need. – DisappointedByUnaccountableMod Dec 26 '20 at 09:24
  • If there was 100% overlap, the area of the merged image would be the same as the area of either of the 2 input images. If there was no overlap at all, the area of the merged image minus the black pixels would equal the sum of the areas of the 2 input images... – Mark Setchell Dec 26 '20 at 10:50
  • the first link is precisely what I would use. keep the stitching transform, but apply it to two white pictures, each in turn. that gives you the mask of each picture in the mapped output. use binary operations (binary and) to get the intersection mask. count pixels (there are functions for that again) to calculate the respective areas. – Christoph Rackwitz Dec 27 '20 at 00:27
  • I have proposed a solution to this in another [question](https://stackoverflow.com/questions/65475791/calculate-percentage-decrease-in-rgb-image) – The Singularity Dec 29 '20 at 08:20

0 Answers0