Multilevel Thresholding Image Segmentation Using Memetic Algorithm

Memetic algorithms (MAs) are hybrid algorithms aimed to improve traditional evolutionary algorithms such as genetic algorithms (GAs). Recently, MAs have been widely used in the image processing field. This paper proposes an image segmentation approach using MA. The proposed approach employs local search as an improvement operator added to the GA in order to speed up the searching process and generate the best solutions faster. The results of the experiments that were conducted on eight different images have shown that MA converges to the solutions faster than GA. In terms of processing time, MA has recorded significant improvement ranging from 17.5% to 79.8%. The comparison results have also shown that MA always achieves the same or better quality of the segmented images.