Singular Value Decomposition in Computer Vision