基于量子粒子群优化的加权平均倒数熵阈值法参数选取  

A weighted average reciprocal entropy thresholding method and its parameter selection based on quantum-behaved particle swarm optimization

作　　者：林爱英[1] 谷小青[1] 李富强[1] 贾树恒[1] 袁超[1] LIN Aiying;GU Xiaoqing;LI Fuqiang;JIA Shuheng;YUAN Chao(College of Science,Henan Agricultural University,Zhengzhou 450002,China)

机构地区：[1]河南农业大学理学院,河南郑州450002

出　　处：《河南农业大学学报》2020年第4期613-619,共7页Journal of Henan Agricultural University

基　　金：国家自然科学基金项目(61703146);河南省科技攻关项目(162102210270);河南省高等学校重点科研项目(16A510018,15A510027)。

摘　　要：考虑到目标和背景对图像分割的不同影响,提出了一种对目标和背景类的倒数熵进行加权平均的阈值分割方法。同时,针对该方法中权重参数m的选取问题,给出一种利用量子粒子群优化算法自适应选取参数的优化策略,依据图像分割质量评价均匀性准则对权重参数m在(0,1)区间进行全局寻优,进而实现了加权平均倒数熵阈值分割方法的自动阈值选取。研究结果表明,该方法可获得比传统熵阈值法更好的分割效果。Considering the different effects of target and background on image segmentation,a threshold segmentation method based on weighted average reciprocal entropy of target and background is proposed.Meanwhile,to solve the problem of how to find the weighted parameter m in the proposed algorithm,an optimized adaptive parameter selecting strategy based on quantum-behavior particle swarm optimization(QPSO)method is proposed.By using an image segmentation quality evaluation criterion named as uniformity measure,the optimal parameter m can be automatically determined in its parameter space(0,1)by the QPSO algorithm,thus realizing the aim of automatic threshold selection using the weighted average reciprocal entropy method.The experimental results show that the proposed method gets better segmentation result than the traditional entropy-based thresholding method.

关 键 词：阈值分割 倒数熵 加权平均 自适应参数选取 

分 类 号：TN911.73[电子电信—通信与信息系统]