Function approximation for medical image registrationFunction approximation for medical image registration Πλήρης Δημοσίευση σε Συνέδριο Conference Full Paper 2019-06-072018enEvolutionary computation has been widely used in intensity-based medical image registration due to its ability to deal with the large number of the local minima which the conventional optimization methods fail. Despite this successful application, they still have certain disadvantages, the most important being the need to do repetitive evaluations of the similarity function for all the candidate solutions, which increases the duration of the image registration process. This disadvantage is more pronounced when the function we seek to optimize is computationally expensive or when the search-space increases due to the large number of degrees of freedom. In this paper, we present a new approximation using a surrogate model for image registration that significantly reduces the time needed for image registration without any quality compromise of the results. The results of the experiments show a decrease of duration up to 40.03%. http://creativecommons.org/licenses/by/4.0/124-12741st International Conference on Telecommunications and Signal Processing Spanakis Konstantinos Σπανακης Κωνσταντινος Mathioudakis Emmanouil Μαθιουδακης Εμμανουηλ Tsiknakis Manolis N. Kampanis, Nikolaos A Marias Kostas Institute of Electrical and Electronics Engineers Function Approximation Harmony Search Machine Learning Mutual Information Rigid Image Registration