decorrelating the eigenvectors..... hi, I have two images which are rotated differently (images are same). I want to align them in one direction (essentially reference free image alignment)...for which I am thinking of taking the help of covariance matrices. I want computer to do this automatically...that is choosing the angle on his own with the help of eigenvectors of the covariance matrices of images. My question is , is it possible if I decorrelate the vectors.? can i find slope of a line if i know the principle eigenvector? will decorrelating the vectors in a given region align them in one direction?