1. Regarding section 9.7.2 Normalized Correlation Metrics
In http://orfeo-toolbox.org/SoftwareGuide/SoftwareGuidech9.html#x32-1660009.7

NC is generally defined by the product of the differences to the mean
divided by the
product of the standard deviations, i.e. see
http://en.wikipedia.org/wiki/Cross-correlation

In http://orfeo-toolbox.org/SoftwareGuide/SoftwareGuidech9.html#x32-1660009.7
it is stated that NC "computes pixel-wise cross-correlation and
normalizes it by the square root of the autocorrelation of the images"
and then the formula indicates that what is actually computed in the
denominator is
sum(A[i]^2), which is *not* equal to autocorr(A[i])

First, the normalization in NC must be by the square root of the
product of the variances, not by the autocorrelation.
Second, for sum(A[i]^2) to be equal to the variance,
you have to previously subtract the mean (aka "centering" in
multi-variate statistics terminology) to both A and B. I think this is
what is actually being done, but it is not stated.

In short:

1.1. The formula is correct, provided the means of A and B have been
previously subtracted (but this fact has to be confirmed and
explicitly stated).
1.2  The definition is wrong. It should be NC "computes pixel-wise
cross-correlation and normalizes it by the square root of the product
of the variances of the images".

2. Now, you understand that there is a lot of ambiguity with what is meant
by the metrics "cross-correlation (CC) and  cross-correlation with
subtracted mean (CCSM)" in the Fine Registration page:
http://www.orfeo-toolbox.org/Applications/FineRegistration.html

3. Note that the engineering jargon differs from the statistical
terminology. In statistics, "Normalized Correlation" is a redundancy,
because correlation is always normalized by definition: the NC formula
is just the Pearson's coefficient.
According to http://en.wikipedia.org/wiki/Cross-correlation, what in
engineering is known as CC is what in statistics is called the sum of
products of deviation (=numerator of Pearson's coefficient).

4. Therefore, my conjectures are:

CC = numerator of Pearson's coefficient
NCC = CCSM =  Pearson's coefficient

Could somebody please confirm or provide the correct answer?

Thanks

Agus

-- 
-- 
Check the OTB FAQ at
http://www.orfeo-toolbox.org/FAQ.html

You received this message because you are subscribed to the Google
Groups "otb-users" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to
[email protected]
For more options, visit this group at
http://groups.google.com/group/otb-users?hl=en
--- 
You received this message because you are subscribed to the Google Groups 
"otb-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to