How to remove false matches in brute force matcher?

asked 2016-12-08 09:53:32 -0600

AJ94 gravatar image

updated 2016-12-09 06:57:13 -0600

My requirement is to compare an image against a set of images in the database and find the best possible match. I have tried using brute force matcher algorithm, but I am getting lot of false matches. Is there a way we can differentiate between true matches and false matches and also calculate the similarity percentage of the images? I am using python.

edit retag flag offensive close merge delete

Comments

1

With brute force matching you can use cross-check, which ensures that each is the best match of the other. This should eliminate many false positives and is very cheap. The alternative is threshold based on how much the best match beats the second best match. If your images are identical and not just potentially similar then there are much simpler approaches than 2D feature matching.

Der Luftmensch gravatar imageDer Luftmensch ( 2016-12-09 06:36:53 -0600 )edit

Thanks for the response. Really appreciate it

AJ94 gravatar imageAJ94 ( 2016-12-10 06:08:09 -0600 )edit