Fuzzy Entropy-based Object Segmentation with an Inertia-Adaptive PSO

Particle Swarm Optimization (PSO) has recently emerged as a simple yet very efficient algorithm for global optimization over continuous spaces. This article describes the application of an improved variant of PSO to the segmentation of objects from complicated real life images. The segmentation task amounts to finding a robust and optimal threshold that separates an object from a background frame. It has been formulated as an optimization problem using the maximum fuzzy entropy principle. Experimentation with several real life images and comparison with the state of the art methods for automatic object segmentation reflect the superiority of the proposed approach in terms of accuracy of the final results and fast computational speed.