摘要
针对现有图像显著性检测算法中显著目标检测不完整和显著目标内部不均匀的问题,本文提出了一种基于多图流形排序的图像显著性检测算法.该算法以超像素为节点构造KNN图(K nearest neighbor graph)模型和K正则图(K regular graph)模型,分别在两种图模型上利用流形排序算法计算超像素节点的显著性值,并将每个图模型中超像素节点的显著值加权融合得到最终的显著图.在公开的MSRA-10K、SED2和ECSSD三个数据集上,将本文提出的算法与当前流行的14种算法进行对比,实验结果显示本文算法能够完整地检测出显著目标,并且显著目标内部均匀光滑.
To resolve the incompletion and non-uniform of salient object detection, this paper proposes an image salient object detection algorithm with multi-graph model and manifold ranking. The algorithm uses superpixels as nodes to construct KNN graph model and K regular graph model. For each model, manifold ranking algorithm is used to calculate saliency values of superpixel nodes. The saliency values of the nodes obtained from the two graph models are fused together with different weights to form the image saliency map. On three public available databases, MSRA-10 K, SED2 and ECSSD, the proposed algorithm is compared with fourteen state-of-art algorithms. Experimental results show that the proposed algorithm can detect a salient object completely, yet the object is uniform and smooth inside.
作者
于明
李博昭
于洋
刘依
YU Ming;LI Bo-Zhao;YU Yang;LIU Yi(School of Computer Science and Engineering,Hebei University of Technology,Tianjin 300401)
出处
《自动化学报》
EI
CSCD
北大核心
2019年第3期577-592,共16页
Acta Automatica Sinica
基金
天津市科技计划(14RCGFGX00846
15ZCZDNC00130
17ZLZDZF00040)
河北省自然科学基金(F2015202239)资助~~
关键词
图像显著性检测
多图模型
流形排序
超像素节点
Image saliency detection
multi-graph model
manifold ranking
superpixel node
