基于三支决策的图像过渡区提取与分割方法  

Method of image transition region extraction and segmentation based on three-way decision

作　　者：杨岚心 苏健民[1] 欧长坤 YANG Lanxin;SU Jianmin;OU Changkun(College of Information and Computer Engineering, Northeast Forestry University, Harbin 150040, China;Institute of Informatics, University of Munich, Munich 80333, Germany)

机构地区：[1]东北林业大学信息与计算机工程学院,哈尔滨150040 [2]慕尼黑大学信息研究所,慕尼黑80333

出　　处：《黑龙江大学自然科学学报》2019年第2期219-226,共8页Journal of Natural Science of Heilongjiang University

基　　金：黑龙江省自然科学基金资助项目(F201028)

摘　　要：在图像分割问题中,过渡区的存在使得分割阈值难以确定,对分割结果的准确性产生较大影响,因此,本文提出了一种基于三支决策的图像过渡区提取与分割方法。建立了图像的粗糙集表示,引入三支决策和最优化理论,把图像分割问题转换为一个在图像域上的分类决策问题,在决策代价取得最小值时,对应的分割阈值即为图像最优分割阈值。该方法综合考虑图像像素分布和分割阈值,设置正则化项。在Weizmann科学研究所的公开图像数据集(Segmentation evaluation database, SED)上取得了81.9%的分割准确率,与其他经典图像分割方法相比,该方法具有更好的分割准确率和稳健性。The segmentation threshold is difficult to be determined because of the existence of the transition region, which has a great influence on the accuracy of segmentation results. An extracting and segmenting method for image transition regions is proposed based on three-way decision. With the rough set representation of images being established, three-way decision and optimization theory are introduced, and the image segmentation problem is transformed into a classification decision-making problem in the image domain. When the minimum decision-making cost is obtained, the corresponding segmentation threshold will become the optimal segmentation threshold. The method considers the image pixel distribution and the segmentation threshold, and sets the regularization term synthetically. Compared with the classical image segmentation methods, our method can achieve the segmentation accuracy rate by 81.9% on the public image dataset ‘Segmentation evaluation database’of Weizmann Research Institute, which owns a better segmentation accuracy and robustness.

关 键 词：图像分割 过渡区 粗糙集 三支决策 最小风险 

分 类 号：Q939.97[生物学—微生物学]