dirtbag's profile - activity

It's mostly to save computation time during RANSAC. If you throw out a bunch of bad matches using a simple distance threshold (3*min_dist), then findHomography has a better chance of finding the right homography quickly. That said, you're idea of only keeping the best K matches would accomplish the same goal of filtering/shortening the list of matches.

Keep in mind that there are images where several keypoints will match perfectly with many, many other keypoints. In this case, you'll need a larger list to be sure that you'll find a good solution. 3*min_dist would then give you the larger list you need while filtering out at least some of the garbage.

I don't have much to add, but here's an image of the difference. The errors look relatively minor. I wonder if it's simply floating point error. That might explain why the errors in angle appear smaller than the errors in magnitude

The documentation has a very nice example of findHomography using RANSAC