局部密度峰聚类耦合字典学习的图像融合算法  

Image fusion algorithm based on local density peak clustering coupled with dictionary learning

作　　者：吴亮 刘国英[2] WU Liang 1,LIU Guo-ying 2(1.College of Software,Anyang Normal University,Anyang 455000,China;2.State Key Laboratory of InformationEngineering in Surveying and Remote Sensing,Wuhan University,Wuhan 430079,Chin)

机构地区：[1]安阳师范学院软件学院,河南安阳455000 [2]武汉大学测绘遥感信息工程国家重点实验室,湖北武汉430079

出　　处：《计算机工程与设计》2018年第7期2008-2014,共7页Computer Engineering and Design

基　　金：国家自然科学基金项目(41001251);河南省重点科技攻关计划基金项目(102102310087);河南省基础与前沿技术研究计划基金项目(112300410182)

摘　　要：为降低当前图像融合算法的冗余信息,提高图像质量,提出基于局部密度峰聚类与字典学习的图像融合方案。将图像划分为若干个图像块,通过信息采样法,选择有用信息的图像块;定义局部密度峰聚类方法,对具有相似结构信息的图像块进行分类,获取不同的图像块簇类;基于K-SVD技术,构建字典学习机制,输出每个簇类的稀疏系数;选择最大值融合准则,对得到的稀疏系数进行融合,获得最终图像。实验结果表明,与当前图像融合方法比较,本文方法的融合质量与鲁棒性更高,其输出融合图像具有更大的边缘强度与相关系数值。To reduce the redundant information of current medical image fusion algorithm and improve the quality of image fusion,a medical image fusion algorithm based on local density peak clustering coupled with dictionary learning was proposed.The image was divided into several image blocks,and the useful information image block was selected using information sampling method.A local density peak clustering method was defined for the classification of image blocks with similar structure information to get different image blocks cluster.The dictionary learning was constructed based on K-SVD,training for each cluster,the sparse coefficients of each cluster class were obtained.The fusion of sparse coefficients was obtained by selecting the maximum fusion criterion for getting the final fusion image.Experimental results show that compared with current medical image fusion methods,the proposed algorithm has higher fusion quality and robustness,the fused image has larger edge intensity and correlation coefficient values.

关 键 词：图像融合 局部密度峰聚类 字典学习 K-SVD方法 信息采样 稀疏系数 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]