APPROXIMATED REPRESENTATION OF IMAGES BY SINGULAR VALUE DECOMPOSITION

Igor Mokriš - Ľubomír Semančík

   The paper deals with approximated representation of images by Singular Value Decomposition (SVD). SVD is based on the computation of singular values of an image matrix. Since the number of singular values can be decreased, the exactness of image representation decreases too, but the error of representation of SVD is minimal in relation to other linear orthogonal transformations. Therefore SVD is a deterministic optimal transformation. As an example, the Slovak banknote in a raster of 408x805 pixels is used and exactness and error of banknote representation is evaluated.

Keywords: singular value decomposition (SVD), approximated representation of images

[full-paper]