A Novel Quantum-Inspired Evolutionary Algorithm for Multi-Sensor Image Registration
Hichem Talbi1, Amer Draa2, and Mohamed Batouche2
1USI Emir Abdelkader, Constantine, Algeria
2Lire Laboratory, Mentouri University, Constantine, Algeria
Abstract: In this paper, we propose a new algorithm for image registration which is a key stage in almost every computer vision system. The algorithm is inspired from both evolutionary algorithms and quantum computing fields and uses the mutual information as a measure of similarity. The proposed approach is based on some concepts and principles of quantum computing such as quantum bit and states superposition. So, the definitions of the basic genetic operations have been adapted to use the new concepts. The evaluation of each solution is performed by the computation of mutual information between the reference image and the resulting image. The process aims to maximize this mutual information in order to get the best affine transformation parameters which allow the alignment of the two images belonging to either similar or different modalities.
Keywords: Image registration, evolutionary algorithms, mutual information, quantum computing, states’ superposition.
Received August 5, 2004; accepted November 3, 2004