基于Contourlet域的隐马尔可夫树模型的图像融合算法  

Image Fusion Algorithm Based on CHMT

作　　者：易小波[1] 龙祖强[1] 邹存[1] 

机构地区：[1]衡阳师范学院计算机系,湖南衡阳421008

出　　处：《电脑与信息技术》2014年第1期11-13,共3页Computer and Information Technology

基　　金：国家自然科学基金项目(项目编号:61074069);湖南省大学生研究性学习和创新项目(项目编号:CX1202)

摘　　要：提出一种基于Contourlet域隐马尔可夫树(CHMT)的多聚焦图像融合方法。CHMT能有效捕获不同尺度系数之间、不同方向系数之间的相关性,能为图像融合提取更多的特征信息。算法对低频子带采用区域方差法,高频子带则依据训练后模型的每一系数的后验概率进行不同的融合处理,以减少融合图像边缘处的斑块模糊现象。仿真实验结果表明,该算法优于基于Contourlet域的常规融合算法,融合后的图像具有更好的主观视觉效果。A image fusion method based on Contourlet_domain Hidden Markov Tree model （CHMT） is proposed .CHMT can capture different scales and directions dependencies of the ＇coefficients. When CHMT is applied to image fusion,the characteristic of images can be effectively extracted and more important information is preserved. In the low-frequency information,the regional variance method is used to obtain the fused low-frequency information. The high-frequency coefficients are modeled using hidden Markov tree and the model is trained to get the posterior probability of the coefficients. According to the posterior probability, different fusion rules are applied. Simulation results show that the algorithm can be effective in reducing the fusion image distortion, and it is an effective image fusion algorithm.

关 键 词：图像融合 CONTOURLET变换 隐马尔可夫树模型 混合高斯模型 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]