A comparative study of nature inspired optimization algorithms on multilevel thresholding image segmentation

In this paper, five successful nature inspired algorithms; the artificial tree algorithm (AT), the particle swarm optimization (PSO), the genetic algorithm (GA), the cultural algorithm (CA), and the cuckoo search algorithm (CS) have been compared on multilevel image thresholding. The segmentation process is based on the Levine and Nazif intra class uniformity criterion which is seen as an optimization problem. The comparison performances are in terms of the value of the objectif function, the peak signal to noise ratio (PSNR) and the computation time. Empirical results over different benchmark images for different threshold numbers reveal the robustness, the reliability and the rapidity of the cultural algorithm (CA).