Research Information System
Engineering Applications of Artificial Intelligence, vol.142, 2025 (SCI-Expanded)
As an optimization problem, the main challenges in the enhancement of dental X-ray images are to perform the tasks of edge detection, noise removal and brightness adjustment in a precise and efficient manner. To overcome these challenges, designing an optimization algorithm that exhibits exploit-exploit behavior in accordance with the geometric structure of the search space of dental images is an important challenge that has not yet been realized. Shortcomings in this area lead to local solution traps and early convergence problems in image enhancement algorithms. This paper introduces the dFDB-LSHADE (dynamic fitness-distance balance-based achievement-history-based adaptive differential evolution with linear population size reduction) algorithm, which is designed for the enhancement of dental X-ray images according to the requirements of the search space of this problem and dynamically changes its exploitation-exploration capabilities. The proposed method is tested on a dataset of 120 periapical images, the most extensive experimental study in the literature. A total of 60 competing algorithms, 53 heuristics and 7 deterministic-based algorithms, were used in the experiments. In the study on dental images, the proposed algorithm has a better Friedman score than all competitors. According to the statistical analysis results obtained from the Wilcoxon pairwise test, the proposed dFDB-LSHADE was able to find better solutions for 23.3% of the images compared to its strongest competitor on the dental image set.