Mastering OpenCV 3 - Second Edition

Eigenvalues, Eigenfaces, and Fisherfaces

Let's view the actual component values in the eigenvalues (as text):

    Mat eigenvalues = model->get<Mat>("eigenvalues"); 
    printMat(eigenvalues, "eigenvalues");

For Eigenfaces, there is one eigenvalue for each face, so if we have three people with four faces each, we get a column vector with 12 eigenvalues sorted from best to worst as follows:

    eigenvalues: 1w18h 1ch 64bpp, range[4.52e+04,2.02836e+06] 
    2.03e+06  
    1.09e+06 
    5.23e+05 
    4.04e+05 
    2.66e+05 
    2.31e+05 
    1.85e+05 
    1.23e+05 
    9.18e+04 
    7.61e+04  
    6.91e+04 
    4.52e+04

For Fisherfaces, there is just one eigenvalue for each extra person, so if there are three people with four faces each, we just get a row vector with two eigenvalues as follows:

 eigenvalues: 2w1h ...