A new heuristic for multilevel thresholding of images

Multilevel thresholding of images helps in separating the interesting objects from their background. The quality of separation depends much on the selected threshold values. This paper proposes a novel thresholding (TH) heuristic for multilevel thresholding problem. Further, the proposed TH heuristic is embedded into whale optimization algorithm (WOA), grey wolf optimizer (GWO), particle swarm optimization (PSO) algorithm to develop new algorithms named WOA-TH, GWO-TH, and PSO-TH respectively. The TH heuristic fine-tunes the best solution of employer nature-inspired algorithms for increasing their capability to escape local optimum. To find the optimum thresholds for an image, between-class variance criterion is employed as the fitness function. Experiments have been performed on twenty benchmark test images using six different number of thresholds. The performance of the proposed algorithms is compared with their respective base algorithms. The results demonstrate that the proposed WOA-TH, GWO-TH, and PSO-TH algorithms are superior to the compared algorithms. Moreover, the computational time of WOA, GWO, and PSO is improved through the incorporation of proposed heuristic.