Multi-Level Image Thresholding Using Modified Flower Pollination Algorithm

Multilevel thresholding is an important approach for image segmentation which has drawn much attention during the past few years. Traditional methods for multilevel thresholding are computationally expensive, because they use the exhaustive searching strategy. To overcome the problem, metaheuristic algorithms are widely applied in this research area for searching the optimal thresholds recently. In this paper, a modified flower pollination algorithm, as a novel improved metaheuristic algorithm, is proposed for multi-level thresholding. Two modifications are proposed to improve the original FPA. First, a fitness Euclidean-distance ratio strategy is employed to modify the local pollination of the original FPA. Second, the global pollination in the original FPA is also biologically modified to improve exploration. Experiments are conducted between seven state-of-the-art metaheuristic algorithms and the proposed one. Both reallife images and remote sensing images are used in the experiments to test the performance of the involved algorithms. The experimental results significantly demonstrate the superiority of our method in terms of the objective function value, image quality measures, and convergence performance.