基于增强教与学优化算法的图像分割  

Research of image segmentation method based on enhanced course TLBO

作　　者：李理 周湘贞[2] LI Li;ZHOU Xiang-zhen(College of Information Engineering,Chengdu Vocational&Technical College of Industry,Sichuan,Chengdu 610213;Department of Information Engineering,Shengda University,Henan,Zhengzhou 451191)

机构地区：[1]成都工业职业技术学院信息工程学院,四川成都610213 [2]郑州升达经贸管理学院信息工程学院,河南郑州451191

出　　处：《贵阳学院学报（自然科学版）》2024年第2期105-109,共5页Journal of Guiyang University：Natural Sciences

基　　金：四川省教育厅项目(项目编号:GZJG2022-060)。

摘　　要：为提高分割图像的质量和提高计算效率,提出了一种基于增强教与学优化算法(ETLBO)的图像分割方法。所提ETLBO利用逆向学习技术,提高全局搜索速度和优化准确度,改进全局寻优性能,并保留了经典课程教与学优化算法计算成本低、稳定性高的优点。此外,采用最小交叉熵(MCE)的概念,将图像分割的多级阈值化问题转换为优化问题,利用ETLBO得到多级最优阈值组合,实现分割图像和原始图像之间的交叉熵最小化,提高分割图像的视觉质量。实验结果表明,所提方法在分割图像的均匀性和适应度方面的性能优于其他先进方法,且计算效率更高。To improve the quality of the segmented images,and the efficiency of computation,an image segmentation method based on enhanced teaching and learning based optimization(ETLBO)is proposed.The traditional TLBO algorithm is improved with the proposed reverse learning technique and small probability mutation strategy,and the global search speed and optimization accuracy are enhanced while retaining the advantages of low computation cost and high stability.In addition,the concept of minimum cross entropy(MCE)is used to transform the thresholding problem of image segmentation into an optimization problem.The optimal combination of multiple thresholds is obtained with ETLBO,which can minimize the cross entropy between the original image and the segmented image,therefore improving the visual quality of the segmented image.The experimental results indicate that the proposed method outperforms other advanced methods in terms of the uniformity and fitness values,and effectively reduces computation time.

关 键 词：图像分割 多级阈值化 最小交叉熵 教与学优化 逆向学习技术 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]