CODEWITHSUNDEEP
c++matlabopencvcross-correlation
Krivers
Krivers

Reputation: 2046

Cross correlation c++ and matlab

I'm trying to do a cross correlation in C++. Problem is that I'm not getting same result in matlab.

C++ code

float telo[5] = {-1, 0, 1, 2, 3};
Mat prueba(1, 5, CV_32F, telo);
float telo2[3] = { 0, 1, 2 };
Mat prueba2(1, 3, CV_32F, telo2);
Mat result;
matchTemplate(prueba, prueba2, result, CV_TM_CCORR);
Result: 2048.0004882812500 2.660783784765e-314#DEN -6.2774385622041925e+066

Matlab code:

a = [-1,0,1,2,3]
b = [0,1,2]
xcorr2(a,b)
Result: -2 -1 2 5 8 3 0

What am I doing wrong?

Upvotes: 2

Views: 2495

Answers (1)

Robert Hegner
Robert Hegner

Reputation: 9376

When I try your C++ code sample with OpenCV 3 I get the following result, as expected:

2, 5, 8

Which is the same as the Matlab solution, but without the tails.

Edit:

To obtain the same result as with Matlab you can add some zero-padding to your input data. Do it manually in your data:

float telo[9] = {0, 0, -1, 0, 1, 2, 3, 0, 0};
Mat prueba(1, 9, CV_32F, telo);

Or a more generic solution (which should also work with 2D data) would be to call

copyMakeBorder(prueba, prueba, prueba2.rows - 1, prueba2.rows - 1, prueba2.cols - 1, prueba2.cols - 1, cv::BORDER_CONSTANT);

before matchTemplate.

Upvotes: 2

Related Questions