Multilevel thresholding based image segmentation using new multistage hybrid optimization algorithm

Thresholding is one of the highly accepted methods for image segmentation because of its simplicity in nature. The selection of optimal threshold values in threshold-based image segmentation is a tricky job. In this work, Kapur’s entropy is used to solve the optimal threshold selection problem and a multistage hybrid nature-inspired optimization algorithm is used to get the best possible parameters for this objective function. The proposed method has three stages namely: primary stage, booster stage and final stage. Particle swarm optimization (PSO), artificial bee colony optimization (ABC) and ant colony optimization (ACO) used at these stages. In this proposed work various benchmarked images have been used for experimentation purpose. The proposed method has been assessed and performance is compared with well-known metaheuristic optimization like PSO, ABC, ACO, classical Otsu thresholding method and modified bacterial foraging optimization qualitatively and quantitatively. Peak signal to noise ratio and Structure Similarity Index are used for qualitative assessment. Wilcoxon p value test, ANOVA test and box plots are used for statistical analysis. The experimental results showed that the proposed method performed better in terms of quality and consistency.