Multilevel minimum cross entropy threshold selection based on particle swarm optimization

Thresholding is one of the popular and fundamental techniques for conducting image segmentation. Many thresholding techniques have been proposed in the literature. Among them, the minimum cross entropy thresholding (MCET) have been widely adopted. Although the MCET method is effective in the bilevel thresholding case, it could be very time-consuming in the multilevel thresholding scenario for more complex image analysis. This paper first presents a recursive programming technique which reduces an order of magnitude for computing the MCET objective function. Then, a particle swarm optimization (PSO) algorithm is proposed for searching the near-optimal MCET thresholds. The experimental results manifest that the proposed PSO-based algorithm can derive multiple MCET thresholds which are very close to the optimal ones examined by the exhaustive search method. The convergence of the proposed method is analyzed mathematically and the results validate that the proposed method is efficient and is suited for real-time applications.