Solving multilevel tresholding problems with artificial bee colony algorithms

Segmentation is an important topic in the image processing. Bi-level segmentation seperating images into two partitions based on a treshold value, multilevel segmentation seperates multiple treshold values. A successful segmentation assigns proper treshold values to optimise a criterion such as a entropy or the between class variance. In this study, tree successful swarm intelligence-based global optimization algorithms, original artificial bee colony (Org-ABC), global best artificial bee colony (G-ABC) and global best-dist artificial bee colony algorithm (GBDIST-ABC) have been employed to find the optimal multilevel tresholds. Two techniques have been used for tresholding. The Kapoor's entropy and between class variance. Experiments have been performed on 10 images using various numbers of tresholds. The results were assessed using statistical tools and suggest that Org-ABC, G-ABC and GBDIST-ABC genarally show equal performance when the number of treshold is two. While the GBDIST-ABC algorithm generally performs better then other techniques when the number of tresholds is greater than two. CPU times of the Org-ABC, G-ABC and GBDIST-ABC algorithms have been reported. These results show us that the algorithms are scalable and running times of the algorithms seem to grow at a linear rate as the problem size increases.