topic Hi Siqing, in Intel® oneAPI Math Kernel Library

Normalized cross-correlation using MKL

Siqing_N_ — Fri, 29 Sep 2017 07:21:27 GMT

I'm trying to use MKL to do NCC in order to find a pattern in a whole image. It seems that vsldCorrExecX only can do correlation but not NCC. for example, pattern = [1 2; 3 4]; image = [1 2; 5 6] We expect to get 0.9762 for sum((pattern-mean(pattern)).*(image-mean(image)))/sqrt((pattern-mean(pattern))^2*(image-mean(image))^2). However, we actually get sum(pattern.*image) = 44. Is there any way to do NCC directly?

Hi Siqing,

Ying_H_Intel — Fri, 13 Oct 2017 05:45:54 GMT

Hi Siqing,

You may consider other two library also: Intel IPP or Intel DAAL

https://software.intel.com/en-us/intel-ipp/documentation,

https://software.intel.com/en-us/intel-daal

Which should provide such functionality.

Best Regards,

Ying