Optimal Threshold Computing in Automatic Image Thresholding using Adaptive Particle Swarm Optimization

Selecting an optimal threshold value is the most important step in image thresholding algorithms. For a bimodal histogram which can be modeled as a mixture of two Gaussian density functions, estimating these densities in practice is not simply feasible. The objective of this paper is to use adaptive particle swarm optimization (APSO) for the suboptimal estimation of the means and variances of these two Gaussian density functions; then, the computation of the optimal threshold value is straightforward. The comparisons of experimental results in a wide range of complex bimodal images show that this proposed thresholding algorithm presents higher correct detection rate of object and background in comparison to the other methods including Otsu's method and estimating the parameters of Gaussian density functions using genetic algorithm (GA). Meanwhile, the proposed thresholding method needs lower execution time than the PSO-based method, while it shows a little higher correct detection rate of object and background, with lower false acceptance rate and false rejection rate.