Computing the Success Factors in Consistent Acquisition and Recognition of Objects in Color Digital Images by Explicit Preconditioning |
| |
Authors: | Maglogiannis Ilias G. Zafiropoulos Elias P. Platis Agapios N. Gravvanis George A. |
| |
Affiliation: | (1) Department of Information and Communication Systems Engineering, University of the Aegean, GR 83200, Karlovasi, Samos, Greece;(2) School of Electrical and Computer Engineering, National Technical University of Athens, GR 15773, Iroon Polytechneiou 9, Athens, Greece;(3) Department of Computer Science, Hellenic Open University, Patras, Greece |
| |
Abstract: | The paper studies the factors influencing the consistent acquisition and recognition of object's color and border features in digital imaging. The proposed image acquisition process is utilized by a computer supported imaging system implementing the acquisition and analysis of skin lesion images supporting medical diagnosis. In addition the same approach may be used for several problems requiring reliable color measurement and object identification. Two methodologies are adopted: The Bayesian Networks, which provide an efficient way of reasoning under uncertainty and are used to incorporate the expert judgement into the estimation of the probability of successful operation, and a Markov chain approach, which is generally used for the dynamic modeling of the system behavior. The Markov chain model requires asymptotically the solution of sparse linear systems. Explicit preconditioned methods are used for the efficient solution of the derived sparse linear system, and the parallel implementation of the dominant computational part is exploited. |
| |
Keywords: | digital image acquisition camera calibration color measurement reproducibility computer vision Bayesian networks Markov modeling Markov Reward Models approximate inverses preconditioning parallel computations |
本文献已被 SpringerLink 等数据库收录! |
|