A Hybrid Method of GA and ACO for 2-D Entropic Automatic Thresholding of Digital Images

Thresholding is an important approach of image segmentation. One of the criteria to select the threshold is two- dimensional (2-D) entropic thresholding. However this method is time-consuming. A hybrid method of GA and ACO is introduced to the 2-D entropic thresholding instead of the exhaustive search. In the proposed method, the ant which represents the threshold vector is at the same time the chromosome. The search space is equally divided into several blocks each marked by pheromone. To reflect the collaboration of ant colony, the 2-D entropy of the ant and the pheromone of its block are both used to construct the fitness function. The best threshold vector is obtained by the genetic evolution of ant colony. Experiments show the proposed method obviously reduced the search time and its accuracy, stability and search efficiency are better than that of the 2-D entropic algorithm by using GA or ACO.