Abstract: In order to describe the texture feature using correlation excellently，a texture feature extraction model according to the weighted Contourlet domain Hidden Markov Tree（CHMT） model is presented，in view of the deficiency of current CHMT model which is just considered the affection of one adjacent node with its child node.This model is considered the information of parent node，as well as the affection of parent node with child node using weight is evaluated，which is reflected through the additional state transition matrix when analyzes the sub node’s status，so that the internal relation of Contourlet coefficient with HMT is described more accurately.At the same time，the similarity is calculated by using K-L distance.The experimental results show that this model is better than the CHMT average retrieval rate of 7%~46%.

Key words: Contourlet domain, Hidden Markov Tree（HMT）, weighted CHMT, texture image, feature extraction

摘要： 为更好地利用相关性描述纹理图像特征，针对目前Contourlet域隐马尔可夫树模型（CHMT）只考虑父结点的一个相邻结点对子结点影响的不足，提出一种加权Contourlet域隐马尔可夫树模型对纹理图像特征提取模型。在分析子结点的状态时，考虑父结点信息的同时利用权重评价父结点兄弟结点对子结点的影响，并通过附加状态转移矩阵体现出来，更加准确地描述了Contourlet系数和HMT的内在联系；运用K-L距离计算图像间的相似度。实验结果表明，改进的模型比CHMT平均检索率高出7%~46%。

关键词: Contourlet域, 隐马尔可夫树, 加权CHMT, 纹理图像, 特征提取