摘要
由于贝叶斯模型和各种图像测量结果,置信传播会更新每个节点的相关概率,提出了在自动交互图像分割过程中应用的新型贝叶斯网络模型。从过度分割模型中的超级像素点区域、边区域、顶点和测量结果之间的统计相关性来构造多层贝叶斯网络模型。除了自动图像分割,贝叶斯网络模型也可用于交互式图像分割中,现有交互分割往往被动地依靠用户提供的准确调整,提出新型主动输入选择方式作为准确调整。实验采用Weizmann数据集和VOC 2006图像集来评估,实验结果表明贝叶斯网络模型可以进行效果更好的自动分割,主动输入选择可以提高整体分割精度。
Bayesian models and a variety of image measurements,belief propagation updates the probability of each node,this paper proposed a new Bayesian network model automatic interactive image segmentation process.From the super pixel area over-segmentation model multilayer Bayesian network model to construct a statistical correlation between the edge of the area,vertex,and measuring results.In addition to automatic image segmentation,the Bayesian network model could also be used for interactive image segmentation that existing interactive segmentation often passively rely on the user to provide accurate adjustment.The new active input selection as a means to accurately adjust this experiment Weizmann datasets and VOC 2006 image sets to assess experimental results show that this Bayesian network model can better automatic segmentation active input selection can improve the overall split accuracy.
出处
《计算机应用研究》
CSCD
北大核心
2013年第4期1240-1243,共4页
Application Research of Computers
基金
湖北省教育厅优秀中青年资助项目(Q20111311)
关键词
超级像素点
贝叶斯网络模型
交互分割
图像分割
super pixel
Bayesian network model
interactive segmentation
image segmentation
