The Most Natural Procedure for Quantum Image Recognition
p. 246-257
Résumé
It is shown how images can be processed, memorized, reconstructed or/and recognized using fundamental and relatively non-artificial quantum dynamics, i.e. | Ψ) =| Ψ)(Ψ | Ψ) in Dirac's notation. The right-most | Ψ) represents the output, the left-most | Ψ) denotes the input, and the central | Ψ)( Ψ | represents the associative memory. No quantum logic gates are needed, but merely a holographic procedure. Our computational model, successfully tested on concrete data, is a quantum version of Hopfield-based neural-netlike associative processing which is mathematicaliy translated into wave-dynamics in a straight-forward way. Here we discuss its most natural quantum implementation(s), i.e. using ordinary interference of image-modulated quantum waves. The non-trivial (even, e.g., anticipatory) capabilities of this model arise from proper consideration of data-structure, or pre-processing by classical systems, therefore it is the best available candidate for the quantum kernel of (conscious) image recognition in the visual cortex.
Texte
Citer cet article
Référence papier
Mitja Peruš et Horst Bischof, « The Most Natural Procedure for Quantum Image Recognition », CASYS, 16 | 2004, 246-257.
Référence électronique
Mitja Peruš et Horst Bischof, « The Most Natural Procedure for Quantum Image Recognition », CASYS [En ligne], 16 | 2004, mis en ligne le 08 August 2024, consulté le 20 September 2024. URL : http://popups.lib.uliege.be/1373-5411/index.php?id=2395
Auteurs
Mitja Peruš
Graz University of Technology, Institute for Computer Vision and Graphics, Inffeldgasse 16, 2.OG, A-8010 Graz, Austria
Horst Bischof
Graz University of Technology, Institute for Computer Vision and Graphics, Inffeldgasse 16, 2.OG, A-8010 Graz, Austria