应用标签域的贝叶斯网络图像分割算法

Application of label domain Bayesian network image segmentation algorithm

通过把概率等级的融合模型和马尔可夫随机场MRF应用于聚类分析模型上来实现图像分割方法,该方法能够更加准确地进行图像分割过程,并最终获得相关融合模型。结合先验概率分布,这种基于能量的Gibbs模型允许指定参数,最大概率等级估计与简单快捷的估计方法进行融合得到分割结果。将此融合框架成功应用在Berkeley图像标准数据库,相关实验结果表明了该方法有效的视觉评价和定量的性能指标,执行结果相比现有分割方法更为突出。

Probability level fusion model and Markov random field are used in the cluster analysis model up to achieve a novel image segmentation method, and this method can be more accurate image segmentation process~ and ultimately fusion model. Prior probability distribution, this energy-based Gibbs model specifies the parameters, the maximum probability level estimated with a simple and efficient estimation method fusion segmentation results. Then fusion framework successfully applied in the Berkeley image standard database related experiments prove that this method~＇~s effective visual evaluation and quantitative per- formance indicators, and compared to the results of the implementation of existing segmentation methods, proposed method is more competitive.