基于精英粒子群优化算法的图像分割方法

AN IMAGE SEGMENTATION METHOD BASED ON ELITE THEORY-IMPROVED PARTICLE SWARM OPTIMIZATION

结合模式识别理论的聚类思想,将基于群智能理论的粒子群优法算法加以改进应用于图像分割中,提出一种基于精英粒子群优化算法的图像分割方法。新方法基于Pareto的精英理论对粒子群算法加以改进,在每次迭代中,生成一个Pareto精英群体,每个粒子更新时的全局极值都是从精英群体中随机选取一个个体作为全局极值。用改进的粒子群优化算法自适应选取分割阈值。实验表明,与遗传算法及标准的粒子群优化算法相比,对于具体的问题该算法具有较好的聚类效果,能够较好地分割图像。

A novel image segmentation method based on elite PSO is provided, which is in combination with clustering thoughts of pattern recognition theory and applies the improved swarm intelligent theory-based particle swarm optimization to image segmentation. The new method derives from the improvement of the PSO with elite theory proposed by Vilfredo Pareto. Every time when iterate, a Pareto elite population is produced, and each global extremum of updating particle is the global extremum of an individual randomly chosen from elite population. The improved particle swarm optimization is employed to adaptively select segmenting threshold. Numerical experiments show that the improved PSO algorithm performs better for the considered problems in clustering result and image segmentation than GA and standard PSO.