摘要
基于图像熵分割复杂图像或多目标图像时,计算量随阈值数呈指数型增长,提出基于改进粒子群人工鱼群(PSO-AF⁃SA)的图像二维熵多阈值快速分割。根据图像二维熵的分割原理,将单阈值分割延伸至多阈值分割。然后定义二维熵多阈值函数为人工鱼群优化算法的目标函数。提出的新算法基于精英策略改进,参考粒子群算法的个体最佳值与群体最佳值,使鱼群算法的寻优能力更加智能高效,在求解二维熵多阈值函数最优值过程中提高收敛速度与寻优精度。最后,对典型图像的分割实验分别与穷举分割法、PSO分割法对比,在单阈值、双阈值和三阈值情况下分别比穷举法快4.51、636、4147.6倍;且收敛速度与寻优精度均优于PSO分割法。实验结果显示:基于改进算法的分割法能更快速精确地解决复杂图像和多目标图像的分割问题。
A fast image segmentation method with multilevel threshold of two-dimensional entropy was proposed on the improved Particle Swarm Opti⁃mization-Artificial Fish Swarm Algorithm(PSO-AFSA)to overcome the large amount of calculation and long computing time of blurred and multi-object complexity images.The single threshold segmentation of two-dimensional entropy was extended to multilevel threshold based on the segmentation principle of image two-dimensional entropy.Then,the bionic optimization algorithm of artificial fish school is in⁃troduced,and the two-dimensional entropy multi-threshold function is defined as the target function of AFSA.The PSO’s individual best value and group best value make the AFSA's bionic optimization ability more intelligent and efficient.Finally,compared with the two-di⁃mensional entropy exhaustive segmentation method and the PSO two-dimensional entropy segmentation method.The algorithm's typical im⁃age segmentation experiments shows that the speed of the proposed method in single threshold segmentation,dual-threshold segmentation and three threshold segmentation are 4.51,636,and 4147.6 times faster than those of the exhaustive method respectly;and the convergence speed and optimization accuracy of the new algorithm are better than the PSO two-dimensional entropy segmentation method.The experi⁃mental results show that the two-dimensional entropy multi-threshold segmentation method based on the improved PSO-AFSA can solve the problem of fuzzy and complex image segmentation more quickly and accurately.
作者
伍蓥芮
张志勇
WU Ying-rui;ZHANG Zhi-yong(Department of Agricultural Engineering,Shanxi Agricultural University,Jinzhong 030801)
出处
《现代计算机》
2021年第1期56-61,共6页
Modern Computer
基金
山西省重点研发计划项目(No.201803D221027-4)
山西省面上自然基金项目(No.201701D121103)。
关键词
图像分割
二维熵
粒子群优化算法
人工鱼群算法
Image Segmentation
Two-Dimensional Entropy
Particle Swarm Optimization
Artificial Fish Swarm Algorithm
