Hello,
I haven't posted any update so far on the progress, so apologies for that. I've been learning OpenCV functions w.r.t the project. The algorithm consists of finding "weak-textured patches" using a gradient covariance matrix. Using the matrix, the eigenvalue and eigenvector can be calculated.
There is a function in OpenCV -- cornerEigenValsAndVecs, which returns a matrix having 2 channels as eigenvalues, plus 4 channels for eigenvectors.
This is where I'm stuck currently. In the paper, the gradient covariance matrix is factored (using SVD), and the resulting singular values are taken as the eigenvalues. I couldn't compare the both, as I seem to be having problems computing the SVD in OpenCV.
_______________________________________________
Digikam-devel mailing list
[hidden email]
https://mail.kde.org/mailman/listinfo/digikam-devel
Locked
