Error in image proximity manual ?

[royvm101]

There seems to be an error in the following link to the manual:

http://software.intel.com/sites/products/documentation/hpc/ipp/ippi/ippi_ch11/ch11_image_proximity_measures.html

According to the manual:
CrossCorrValid_NormLevel
Computes normalized cross-correlation between an image and a template.

However, should it be the normalized cross-correlation coefficient ? (the word coefficient is missing, and its addition means a totally different algorithm ?)

[igorastakhov]

IPP solution is based on http://scribblethink.org/Work/nvisionInterface/nip.html (with several modifications oriented to performance optimization based on Intel architecture specific) - you can see that both wordings are used in this article -normalized cross-corelation as well as normalized cross-correlation coefficient. As result of IPP functionis not 1 single coefficient but a matrix of such coefficients - we prefer to use "normalized cross-corelation" term.

Regards,
Igor

[royvm101]

But in the link of the manual I provided in my original message there is inconsistency:

• CrossCorrFull_NormLevel
  Computes normalized correlation coefficient between an image and a template.
• CrossCorrSame_NormLevel
  Computes normalized correlation coefficient between an image and a template.
• CrossCorrValid_NormLevel
  Computes normalized cross-correlation between an image and a template.

The first two use the word coefficient, and the last one does not, even though the only difference between the three options are the Full/Same/Valid options.

[igorastakhov]

Ok, thank you, will bring our documentation to consistent state in the next 8.0 release.

regards,
Igor