阈值分割方法的关键在于阈值选取。阈值决定了图像分割结果的好与坏,随着阈值数量的增加,图像分割的计算过程越来越复杂。为了选取适当的阈值进行图像分割,文中提出了离散灰狼算法(Discrete Grey Wolf Optimizer,DGWO),即经过离散化处理的灰狼算法,并用该算法求解以Kapur分割函数为目标函数的全局优化问题。 DGWO算法具有很好的全局收敛性与计算鲁棒性,能够避免陷入局部最优,尤其适合高维、多峰的复杂函数问题的求解,并且可以很好地融合到图像分割过程当中。大量的理论分析和仿真实验的结果表明,与遗传算法( GA)、粒子群算法( PSO)的图像分割结果相比,在选取多张分割图像、多个分割阈值的情况下,该算法具有更好的分割效果,更高的分割效率,优化得到的阈值范围更加稳定,分割质量更高。%The key of threshold segmentation is to select the thresholds which can determine the result of segmentation. With the increas-ing amounts of thresholds,the computation complexity gets higher. In this paper,a Discrete Grey Wolf Optimization ( DGWO) is pro-posed to select the appropriate thresholds for image segmentation and apply it to the global optimization problem of objective function of Kapur segmentation function. The DGWO can be well blended into image segmentation. It specially suits for solving complex function with high-dimension and multi-peak for its excellent performance in global convergence,robustness and ability to avoid trapping into lo-cal optimization. Extensive theoretical analysis and the results of simulation have shown that DGWO has better effectiveness,efficiency, stability of the range of thresholds and quality in multi-images and multi-thresholds segmentation compared with GA and PSO.
展开▼
