Utilisation de l'intelligence en essaim pour le traitement d’image

Abstract

In this study, we propose an image segmentation approach based on swarm intelligence,specifically using the Grey Wolf Optimizer (GWO) algorithm. First, a single-threshold segmentation is performed using Otsu’s ethod, and the resulting segmented image serves as a reference (ground truth) to evaluate the performance of GWO. The thresholds obtained by Otsu and GWO are 154 and 156, respectively, with a fitness value reaching 0.998, indicating a strong similarity. Single-threshold segmentation does not present major difficulties for either approach. However, multi-threshold segmentation (particularly with three thresholds) poses a significant hallenge for Otsu due to its brute-force exhaustive method, which becomes computationally intractable as the number of possible combinations increases. Our approach is tested on a set of ten images of various sizes. For each image, the result of Otsu’s method is used as ground truth to evaluate the segmentation produced by GWO. Experimental results demonstrate the promising performance of the GWO algorithm, offering a good trade-off between segmentation accuracy and computational efficiency.