So I rejoiced in the easy installation of opencv under WSL2 (Windows Subsystem for Linux), Ubuntu:
apt-get install libopencv-dev
Following this up with dpkg -s libopencv-dev | grep Version answered me
Version: 4.5.4+dfsg-9Ubuntu4. I am interested in eigen vectors.
Nothing evil, real valued eigen values will do nicely (however, complex ones would also be welcome).
I wrote a CMakeLists.txt
cmake_minimum_required(VERSION 3.5)
project(testcv)
find_package( OpenCV REQUIRED )
include_directories(${OpenCV_INCLUDE_DIRS})
add_executable(testcv testcv.cpp)
target_link_libraries(testcv ${OpenCV_LIBS})
And the following test program testcv.cpp
#include <iostream>
#include <opencv2/core/mat.hpp>
#include <opencv2/imgproc.hpp>
using namespace cv;
using namespace std;
int main()
{
float numbers[] =
{ 0, 2,
1, 0 };
Mat A = Mat(2,2,CV_32F,numbers);
Mat eigen_vals, eigen_vecs;
eigen(A, eigen_vals, eigen_vecs);
cout << "A: " << A << endl <<
"eigen_vals: " << eigen_vals << endl <<
"eigen_vecs: : " << eigen_vecs << endl <<
"Done." << endl;
return 0;
}
Finally, saying
cmake .
make
./testcv
got me
A: [0, 2;
1, 0]
eigen_vals: [0.99999994;
-0.99999994]
eigen_vecs: : [0.70710677, 0.70710677;
-0.70710677, 0.70710677]
Done.
Honestly? That is just wrong! For comparison, my octave results:
>> A
A =
0 2
1 0
>> lambda
lambda =
Diagonal Matrix
1.4142 0
0 -1.4142
>> B
B =
0.8165 -0.8165
0.5774 0.5774
>> A*B
ans =
1.1547 1.1547
0.8165 -0.8165
What am I doing wrong here, in such a simple call to such a simple interface?
Found my own answer by reading the docs a little closer: https://docs.opencv.org/4.x/d2/de8/group__core__array.html
bool cv::eigen (InputArray src,
OutputArray eigenvalues, OutputArray eigenvectors=noArray())
Calculates eigenvalues and eigenvectors of a symmetric matrix.
Also answering the optional "complex" part of the question (real-valued, symmetric matrices yield real-valued eigenvalues).
Note: Also be wary that eigenvectors are stored row-wise by cv::eigen(..). So the first row is the first eigenvector.