基于参数初始化Contourlet HMT模型的图像分割
Image Segmentation Using Contourlet Domain Hidden Markov Model Based on Parametric Initialization
基于Contourlet的多尺度和多方向性，本文提出了一种改进的Contourlet域HMT模型的图像分割方法(Improve Contourlet-domain hidden markov tree model, ICHMT)。我们采用一种参数初始化规则对训练图像进行HMT模型的参数初始化估计，并结合上下文模型JMCMS(a Joint Multicontext and Multiscale)对CHMTseg(Contourlet-domain hidden markov tree image segmentation)方法进行改进。为说明本方法的有效性，我们分别对纹理图像和SAR图像进行与CHMTseg方法的对比实验。实验结果表明我们的分割方法对粗分割的分割质量有较好的改良作用，并能在多尺度融合中更充分的利用多背景信息得到较好的分割效果。 实验结果证明了方法的有效性。
Based on the multiscale and multidirectional characteristics of Contourlet coefficient, this paper improves a contourlet-domain hidden markov tree model for image segmentation. We utilize an initialization scheme for parametric of hidden markov tree model and combine the context model JMCMS (a Joint Multicontext and Multiscale) with CHMTseg (Contourlet-domain hidden markov tree image segmentation) to obtain our method. In order to evaluate the performance of the proposed method, we make experiments in synthetic mosaic image and SAR image. The segmentation results were compared with CHMTseg. Experiment results show that our method has better simulation results than CHMT in coarse segmentation and have better effect in homogeneous regions reservation.
侯彪、徐婧、焦李成
计算技术、计算机技术
图像分割HMT参数初始化JMCMS
Image segmentationCHMTparametric initializationJMCMS
侯彪,徐婧,焦李成.基于参数初始化Contourlet HMT模型的图像分割[EB/OL].(2009-01-09)[2025-07-16].http://www.paper.edu.cn/releasepaper/content/200901-358.点此复制
评论