树状小波与多尺度估计理论的图像融合技术  

An Image Fusion Algorithm based on Tree Structure Wavelet and Multi-resolution Estimation Theory

作　　者：同武勤[1] 杨华[1] 金伟[1] 杨旭[2] 

机构地区：[1]电子工程学院安徽省红外与低温等离子体重点实验室 [2]77680部队

出　　处：《火力与指挥控制》2009年第8期44-48,共5页Fire Control & Command Control

基　　金：安徽省重点实验室基金资助项目

摘　　要：针对红外和可见光图像的特点及相互融合的应用,提出一种基于树状小波多尺度估计理论的红外与可见光图像融合算法。首先对源图像进行树状小波多尺度分解,依据子图像的信息量得到塔式结构子图像;然后基于不同子图像对应层上的对应像素,根据EM算法估计模型参数,采用SAGE迭代算法优化估计参数得到融合子图像;最后根据小波逆变换获得融合图像。实验结果表明:该融合算法能够更好地综合利用红外图像较好的目标指示特性与可见光图像较清晰的场景信息;性能评估显示:该算法得到的融合图像互信息较文献[5,6]分别提高了49.60%和24.90%,均方根误差分别减小了66.40%和56.00%。According to the characteristic and application of IR and visual images, a new fusion algorithm is presented to IR and visual image based on tree structure wavelet and multi-resolution estimation theory. Firstly, the images are decomposed by the tree structure multi-resolution wavelet, and the pyramidal structure sub-images are acquired under information energy. Secondly, the fusion sub- images are obtained by SAGE parameters estimation algorithm, all which base on the relevant pixel of different sub-images and the EM algorithm. Finally, the fusion image is acquired through wavelet anti- transformation. The experiment shows this algorithm can integrate better target display character of IR image and clearer scene information of visual image. The performance evaluation shows the MI respectively increases 49.60%and24. 90% and the RMSE respectively decreases 66.40% and 56.00% than the reference [5] and [6].

关 键 词：统计模型 估计理论 期望值最大 性能评估 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]