摘要
简单线性迭代聚类算法(SLIC)作为目前主流的基于聚类的超像素分割算法,能产生形状规整的超像素,但是边界附着度不高,针对以上问题本文提出了基于综合辨识信息的SLIC超像素分割算法。该算法首先调整种子点的初始化选取方式,计算像素梯度值,扩大初始聚类中心的选取范围。其次在距离度量时,加入像素的边缘概率,以权重的方式加入到距离公式中,减少了像素的误分割现象。实验结果表明,本文方法与SLIC算法相比,在分割质量方面有明显提升;同时与其他几种算法相比,本文提出的算法可以有效地提高超像素的边界附着度,同时降低像素的分割错误率。
Simple linear iterative clustering(SLIC),as the current mainstream clustering-based superpixel segmentation algorithm,can produce superpixels with regular shapes,but the boundary adhesion is not high.In view of the above problems,a comprehensive SLIC super pixel segmentation algorithm for identifying information is proposed.The algorithm first adjusts the initial selection method of seed points,calculates the pixel gradient value,and expands the selection range of the initial cluster center.Secondly,when measuring the distance,the edge probability of the pixel is added,and it is added to the distance formula in the way of weight,which reduces the phenomenon of mis-segmentation of the pixel.The experimental results show that compared with the SLIC algorithm,the method in this paper has a significant improvement in segmentation quality.At the same time,compared with several other algorithms,the proposed algorithm can effectively improve the boundary attachment of superpixels,and at the same time reduce the segmentation error rate of pixels.
作者
陈莹莹
康艳
李文法
宏晨
Chen Yingying;Kang Yan;Li Wenfa;Hong Chen(Smart City College,Beijing Union University,Beijing 100101;College of Robotics,Beijing Union University,Beijing 100101)
出处
《高技术通讯》
CAS
2021年第8期816-823,共8页
Chinese High Technology Letters
基金
国家自然科学基金(61972040,61502475)
北京联合大学研究生科研创新(YZ2020K001)资助项目。
关键词
超像素分割
聚类
简单线性迭代聚类(SLIC)
辨识信息
superpixel segmentation
clustering
simple linear iterative clustering(SLIC)
identification of information
