应用耦合对象相似度的阈值分割方法研究  

Threshold Segmentation Algorithm Research Used on Coupled Object Similarity

作　　者：武玉坤[1] WU Yukun(School of Management and Information,Zhejiang Post and Telecommunication College,Shaoxing 312016)

机构地区：[1]浙江邮电职业技术学院管理与信息学院,绍兴312016

出　　处：《计算机与数字工程》2017年第6期1194-1199,共6页Computer & Digital Engineering

基　　金：国家自然科学基金青年项目(编号:61602059);全国教育信息技术规划课题(编号:126240629)资助

摘　　要：传统Otsu算法及其改进算法将类间方差设定成最优阈值,从而使得针对直方图分布区别的图像分割效果产生较大区别,论文提出应用耦合对象相似度来进行阈值分割的改进方法来解决。首先,构建模型描述耦合对象相似度,模型能够综合考虑各种对象属性及属性之间关联,以高准确度和低复杂度来描述耦合对象关系;其次,应用耦合对象相似度来替代传统Otsu算法的类间方差作为新条件,将所选阈值划分成每个类看成是耦合对象相似度模型中的对象,每个类都有概率和灰度均值两种属性,通过计算类间相似度并在类间相似度最小时获取最优阈值。实验结果表明,应用耦合对象相似度执行阈值分割算法能够有效提高描述类间差异精确度和图像分割效果,对于单分布与双峰显著且底部平坦的特征图像具有较强适应能力。Since the Otsu method and most of its improved methods take between-class variance as the foundation of pickingthreshold and the great difference of image segmentation results for different histogram distribution images,a new threshold segmen?tation algorithm which based on coupled object similarity is proposed in this paper.Firstly,a model of coupled object similarity is in?troduced,which can take both the relationship of the various attributes of the objects itself and the relationship between the proper?ties into account,and can capture the relationships between the objects with high accuracy and low algorithm complexity.Secondly,between-class variance in the Otsu method is replaced by coupled object similarity to pick threshold,each class distinguished bythe selected threshold is regarded as an object in the model of coupled object similarity,each class has two attributes,the probabili?ty of class and gray mean.Similarity between classes is calculated,and the optimal threshold value is obtained according to the mini?mum of similarity between classes.The experiments can prove that the proposed algorithm can measure the difference of classes at ahigher accuracy and obtain better segmentation results.

关 键 词：图像分割 类间方差 耦合对象相似度 类间相似度 最优阈值 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]