[1]袁恒东.基于标签传递图割的图像分割算法[J].计算机技术与发展,2017,27(12):35-38.[doi:10.3969/ j. issn.1673-629X.2017.12.008]
　YUAN Heng-dong.An Image Segmentation Algorithm Based on Label Propagation Graph Cut[J].Computer Technology and Development,2017,27(12):35-38.[doi:10.3969/ j. issn.1673-629X.2017.12.008]
点击复制

基于标签传递图割的图像分割算法()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
27
期数:
2017年12期
页码:
35-38
栏目:
出版日期:
2017-12-10

文章信息/Info

Title:
An Image Segmentation Algorithm Based on Label Propagation Graph Cut
文章编号:
1673-629X(2017)12-0035-04
作者:
袁恒东
南京理工大学 计算机科学与工程学院,江苏 南京 210094
Author(s):
YUAN Heng-dong
School of Computer Science and Engineering,Nanjing University of Science and Technology,Nanjing 210094,China
关键词:
交互式图像分割; 图割; 超像素; 标签传递
Keywords:
interactive image segmentation; graph cut; super pixels; label propagation
分类号:
TP301.6
DOI:
10.3969/ j. issn.1673-629X.2017.12.008
文献标志码:
A
摘要:
基于图割的交互式图像分割方法实现了基于用户感兴趣区域的图像前景和背景分割,在计算机视觉和图像处理领域受到了众多研究者的关注。 传统图割算法往往是基于图像局部特征进行分割,其收敛速度以及图像结构描述能力具有一定的局限性。 为了进一步提高分割精度,提出一种基于标签传递图割的交互式图像分割算法。 首先通过引入三层超像素层构造图模型,实现高层信息的计算。 通过多层超像素层,既能抑制过分割,也可以通过不同参数的超像素来增加算法的鲁棒性,提升算法稳定性能。 然后结合高次信息,采用已标记样本特征指导未标记样本的标签分配,进行标签传递,利用较少标记样本对未标记样本进行聚类,从而提高了学习精度。 最后使用最大流/ 最小割算法求解最终的分割结果。在SRC 和 Berkeley 数据集上进行的分割实验表明,基于标签传递和图割的图像分割算法是有效的。
Abstract:
The interactive image segmentation algorithm based on graph cut segments the foreground and background in the image based on the region of interest of the users,and has received the attention of many researchers in the field of computer vision and image processing. Traditional image segmentation algorithm is usually based on local features of image,which is limited to its convergence speed and description of image structure. In order to further improve the segmentation accuracy,an interactive image segmentation algorithm based on label propagation and graph cut is proposed. Firstly,a three-layer super-pixel layer structure graph model is introduced to consider the high-level information,which can further improve its robustness and stability. Then,the label propagation technology is utilized to cluster the unlabeled samples with limited labeled samples and improve the segmentation accuracy by combining the local and high-order information. Finally,the maximum flow/ minimum cut algorithm is used to achieve the final segmentation result. Experiments on MSRC and Berkeley datasets demonstrate the effectiveness of the proposed algorithm comparing with state-of-the-art methods.

相似文献/References:

[1]彭智东,宣士斌.融合最小生成树和四叉树的图割图像分割方法[J].计算机技术与发展,2018,28(12):102.[doi:10．3969/j． issn．1673－629X．2018．12．022]
　PENG Zhidong,XUAN Shibin.Image Segmentation Method Based on Graph Cut Combining Minimum Spanning Tree and Quadtree[J].Computer Technology and Development,2018,28(12):102.[doi:10．3969/j． issn．1673－629X．2018．12．022]
[2]汪 昊,刘向阳.基于区域中心的交互式图像前景提取方法[J].计算机技术与发展,2020,30(02):12.[doi:10. 3969 / j. issn. 1673-629X. 2020. 02. 003]
　WANG Hao,LIU Xiang-yang.An Interactive Image Foreground Extraction Method Based on Region Center[J].Computer Technology and Development,2020,30(12):12.[doi:10. 3969 / j. issn. 1673-629X. 2020. 02. 003]
[3]孙凯月,刘向阳.基于非均匀热扩散的交互式图像分割算法[J].计算机技术与发展,2021,31(03):84.[doi:10. 3969 / j. issn. 1673-629X. 2021. 03. 014]
　SUN Kai-yue,LIU Xiang-yang.Interactive Image Segmentation Algorithm Based on Non-uniform Heat Diffusion[J].Computer Technology and Development,2021,31(12):84.[doi:10. 3969 / j. issn. 1673-629X. 2021. 03. 014]
[4]陈丹冉,王 健.基于图割和局部算子的图子集选取[J].计算机技术与发展,2023,33(06):1.[doi:10. 3969 / j. issn. 1673-629X. 2023. 06. 001]
　CHEN Dan-ran,WANG Jian.Graph Subset Selection via Graph Cut and Localization Operators[J].Computer Technology and Development,2023,33(12):1.[doi:10. 3969 / j. issn. 1673-629X. 2023. 06. 001]

更新日期/Last Update: 2018-03-05