Zero-normalised cross correlation for Template Matching

asked 2020-03-01 11:03:18 -0600

Maksat gravatar image

I want to use Template Matching for tracking some objects and some literature says Zero-normalised versions of Cross Correlation (ZNCC) or Sum of Square Differences (ZNSSD) are the most robust. I would really appreciate if anyone could advise if Template Matching includes these methods of correlation criteria and how. Thank you

edit retag flag offensive close merge delete

Comments

matchtemplate doc and method parameters here

LBerger gravatar imageLBerger ( 2020-03-02 02:17:54 -0600 )edit

I want to use Template Matching for tracking some objects

that sounds like a bad idea, far too much overhead. rather implement your own cross-correlation using your desired normalization, maybe take a look at the MOSSE tracker-

berak gravatar imageberak ( 2020-03-02 02:37:56 -0600 )edit