基于改进拉普拉斯能量和的快速图像融合  

Fast image fusion based on sum of modified Laplacian

作　　者：孙晓龙[1] 王正勇[1] 符耀庆[2] 易云[1] 何小海[1] 

机构地区：[1]四川大学电子信息学院,成都610064 [2]中海油能源发展股份有限公司北京分公司,北京100027

出　　处：《计算机工程与应用》2015年第5期193-197,共5页Computer Engineering and Applications

基　　金：国家自然科学基金(No.61071161)

摘　　要：为了得到优质的融合图像,提出了一种改进的拉普拉斯能量和(New Sum of Modified Laplacian,NSML)多聚焦图像融合算法。该算法在传统SML计算每个像素点的变步长拉普拉斯算子值仅有的水平和垂直方向的基础上,增加了斜对角线上的四个方向。同时通过分析NSML算法的计算过程,发现存在大量的重复计算,从而提出了基于积分图像的快速NSML图像融合方法。该方法通过简化NSML的计算过程,大大减少了图像融合处理过程消耗的时间,提高了图像融合的效率。实验结果表明,快速NSML方法在达到极佳融合图像质量的同时,提升了算法的实时性。A new algorithm is proposed based on Sum of Modified Laplacian（NSML）, for the fusion problem of the multi-source images of the same scene. The traditional SML only calculates the Laplacian value of each pixel on horizontal and vertical direction, while NSML adds the four directions on the diagonal ramp. Fast NSML based on integral image is put forward in order to solve the problem that there are amounts of repeated calculations in the process of NSML. Fast NSML greatly reduces the time consumed by the fusion process and improves the efficiency of fusion by mean of simplifying the calculation process. The experimental results show that the method achieves excellent fusion of image quality, at the same time, enhances the real-time algorithm.

关 键 词：图像融合 改进拉普拉斯能量和 空间域 一致性校验 积分图像 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]