Image Matching using Cross Correlation (Cyrill Stachniss, 2021)

[@huangyou535]

I am a graduate student from Taiwan. This video is very helpful for me. Thank you so much!

[@leecheng2005]

An outstanding lecture on template matching both in theoretical and in pratical.

[@naderrazmara8562]

This is one of the best explanation about cross correlation and its use in image procwssing.

[@CyrillStachniss]

Thanks! 😀

[@naderrazmara8562]

Thanks a lot. It is one of the best discription about crosscorrelation in template matching

[@ramanabotta6285]

Thank you professor

[@lsc1356]

Thank you professor, this lecture is very helpful, professor, I have a question, When image matching using cross correlation in the Fourier domain, why is there a strong signal from the top-left? I'd really appreciate it if you could let me know.

[@arjun9273]

why do we have a factor 1/(M-1) instead of having 1/M in that Cross correlation formulae ?

[@CyrillStachniss]

It comes from the estimation of the covariance

[@dipakkambale8187]

@@CyrillStachniss It should be 1/M. The variance is defined as mean of square of deviation from mean...

[@MichaelMillard-b5k]

@@CyrillStachniss dividing by 1/(M-1) is used when computing the sample variance to account for bias; 1/M is used when computing the entire population variance. Since we are considering every single pixel in the template here (i.e. the entire population), should we not use the latter formula (divide by 1/M)?

[@ЗакировМарат-в5щ]

good

[@guccimyerz8397]

thx

[@navinbondade5365]

Coding video