多聚焦图像融合中最优双树复小波分解层数的选择

摘要

采用具有近似平移不变性和方向选择性的双树复小波变换对多聚焦图像进行多分辨率分析与重构，是一种高效的融合方法。然而，分解层数的选择是影响该算法性能的重要因素之一。本文提出了一种基于信息熵、互信息量、边缘融合质量和加权融合质量等融合质量评价指标确定最优分解层数的方法。实验结果表明，不同的原始图像用同种融合算法分解重构，最优分解层数不一定相同；相同的原始图像用不同的融合算法分解重构，最优分解层数也不一定相同。因此，只有综合考虑多种代表性的评价指标，才能确定最优分解层数。%The dual-tree complex wavelet transform,with near shift-invariance and directionality,is a very effective method used in multi-focus image fusion. However,selection of decomposition level is one of the important factors affecting the performance of algorithm.This paper presents a method to determine the optimal decomposition level based on fusion quality evaluation index,such as entropy,mutual information,edge-dependent fusion quality index,weighted fusion quality index.The experiment result show that, different original image with the same fusion algorithms,the optimal decomposition level is not necessarily the same.The same original image with different fusion algorithms,the optimal decomposition level is also not necessarily the same.Therefore,only considering sevral representative evaluation index can you determine the optimal decomposition level.