Optimal algebraic coding of noisy and distorted images

The traditional approach to coding of noisy and distorted images assumes preliminary enhancement and restoring of images and only then in fact their coding. A novel technique is proposed, where the indicated procedures are combined in optimum algebraic coding. At the heart of the approach is the optimization of coordinate basis of an image representation. It is shown, that the required optimum properties have the eigenfunctions of Fisher's information operator known in the statistical theory of estimation. These functions satisfying defined criteria of a selection are called Principal Information Components (PIC). The PIC-projection technique for improvement of an arbitrary image decomposition is proposed. The given technique allows primary physical representation of an image to decrease a statistical error of its coding on noisy data.