一种基于非均匀马尔可夫随机场的图像分割方法  

A image segmentation algorithm based on inhomogeneous Markov random field

作　　者：张辉[1] 胡阳涟[2] 

机构地区：[1]咸阳师范学院数学与信息科学学院,陕西咸阳712000 [2]西安理工大学理学院,陕西西安710054

出　　处：《电子设计工程》2012年第17期146-149,共4页Electronic Design Engineering

基　　金：陕西自然科学基金(2010JQ1013);陕西省教育厅专项科研基金(2010JK896)

摘　　要：提出了一种新的基于非均匀马尔可夫随机场(MRF)的图像分割算法。基于非均匀马尔可夫随机场的图像分割的关键是对MRF中耦合系数的估计。本文结合四叉树分解提出了一种新的非均匀MRF的耦合系数估计方法。先对图像用传统的MRF分割方法进行预分割,再在预分割的基础上用边缘检测算子检验出预分割图像中的边缘,再利用图像的边缘信息对图像进行四叉树分解,把图像分成不同大小的子块。再根据每个子块的大小,估计出非均匀MRF的耦合系数。实验表明,将本文方法估计出来的耦合系数应用到分割算法中去,能明显改善图像分割的效果,而且具有更好的自适应性。A novel image segmentation algorithm based on inhomogeneous markov random field （IMRF） is proposed in this paper. The key technique in image segmentation algorithm based on IMRF is the estimation of the isomorphic coefficient in IMRF. A novel estimate method based on quad-tree decomposing is proposed in this paper. After the pre-segmentation of the image based on the traditional algorithm based on IMRF, an edge detection was done with the image. And then a quad-tree decomposing was done use the edge information. So the image was decomposed into different sizes of sub-blocks. Each size has it＇s own isomorphic coefficient. The experiment results indicated that the isomorphic coefficient estimated by this algorithm can significantly improve the effectiveness of the segmentation ,and also has more adaptability.

关 键 词：非均匀马尔可夫随机场 耦合系数 四叉树分解 自适应 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]