利用混沌布谷鸟优化的二维Renyi灰度熵图像阈值选取  

Two-dimensional Renyi-gray-entropy image threshold selection based on chaotic cuckoo search optimization

作　　者：马英辉[1,2] 吴一全[1,3,4,5] 

机构地区：[1]南京航空航天大学电子信息工程学院,江苏南京211106 [2]宿迁学院信息工程学院,江苏宿迁223800 [3]西华大学制造与自动化省高校重点实验室,四川成都610039 [4]华中科技大学数字制造装备与技术国家重点实验室,湖北武汉430074 [5]安徽理工大学煤矿安全高效开采省部共建教育部重点实验室,安徽淮南232001

出　　处：《智能系统学报》2018年第1期152-158,共7页CAAI Transactions on Intelligent Systems

基　　金：西华大学制造与自动化省高校重点实验室开放课题(S2jj2014-028);华中科技大学数字制造装备与技术国家重点实验室开放课题(DMETKF2014010);安徽理工大学煤矿安全高效开采省部共建教育部重点实验室开放课题(JYBSYS2014102)

摘　　要：为了进一步降低现有的Renyi熵阈值法的计算复杂度,提出了基于混沌布谷鸟算法和二维Renyi灰度熵的阈值选取。首先,引入一维Renyi灰度熵阈值选取公式,建立基于像素灰度和邻域梯度的二维直方图,推导出基于该直方图的二维Renyi灰度熵阈值选取公式,通过快速递推公式来减少阈值准则函数的计算量;最后,采用混沌布谷鸟算法搜索最优阈值来完成图像分割。结果表明,与二维Arimoto熵法、基于粒子群的二维Renyi熵法、基于混沌粒子群的二维Tsallis灰度熵法、基于布谷鸟算法的二维Renyi灰度熵法相比,所提出的方法能够准确实现图像分割,且运算速度有所提升。To further reduce the computational complexity of existing thresholding methods based on Renyi＇s entropy,in this paper, we propose a method for threshold selection based on 2-D Renyi-gray-entropy image threshold selection and chaotic cuckoo search optimization. First, we derive the formula for a 1-D Renyi-gray-entropy threshold selection.Then, we build a 2-D histogram based on the grayscale and gray-gradient and derive a formula for 2-D Renyi-gray-entropy threshold selection based on this histogram. We use fast recursive algorithms to eliminate redundant computation in the threshold-selection criterion function. Finally, to achieve image segmentation, we search for the optimal threshold using the chaotic cuckoo search algorithm. The experimental results show that, compared with 2-D Arimoto-entropy thresholding method, the 2-D Renyi-entropy thresholding method based on particle swarm optimization, the 2-D Tsallisgray-entropy thresholding method using chaotic particle swarm, and the 2-D Renyi-gray-entropy thresholding method based on the cuckoo search, our proposed method can segment objects more accurately and has a higher running speed.

关 键 词：图像分割 阈值选取 布谷鸟算法 Renyi灰度熵 灰度-梯度二维直方图 混沌优化 Arimoto熵 Tsallis灰度熵 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]