Inna Kozlov and Vladimir Ovchinnikov: Optimal feature space for representation of finite dimensional data

HAIT Journal of Science and Engineering C
Volume 4, Issues 1-2, pp. 194-202
© 2007 Holon Institute of Technology

Optimal feature space for representation
of finite dimensional data

Inna Kozlov^1,* and Vladimir Ovchinnikov ^2,3

¹Holon Institute of Technology, 52 Golomb Str., Holon 58102, Israel
²Mathematical Institute, Academy of Science of the Czech Republic,
Zitna 25, CZ-115 67 Praha 1, Czech Republic
³Mathematical Department, Voronezh State University,
Universitetskaia pl. 1, Voronezh 394006, Russia
^*Corresponding author: innak@hait.ac.il
Received 5 April 2005, accepted 5 December 2005

The paper develops a recent Coifman's approach to the problems of Signal and Image Processing, based on spectral theory of operators in Hilbert spaces. New notion of canvas is introduced. This allows to use powerful tools of Operator Ideal Theory which, in turn, leads to a numerical algorithm for restoring the original image.

Full text PDF (153 KB)

Optimal feature space for representation of finite dimensional data

Inna Kozlov1,* and Vladimir Ovchinnikov 2,3

Optimal feature space for representation
of finite dimensional data

Inna Kozlov^1,* and Vladimir Ovchinnikov ^2,3