基于滑窗方差和形态学的多聚焦图像融合  

Multi-focus Image Fusion Based on Sliding Window Variance and Morphology

作　　者：罗回彬[1,2] 林梓鑫 怀国威 温焕锦 Luo Huibin;Lin Zixin;Huai Guowei;Wen Huanjin(Beijing Institute of Technology,Zhuhai 519000;Faculty of Information Technology,Macao University of Science and Technology,Macao 999078,China)

机构地区：[1]北京理工大学珠海学院,珠海519000 [2]澳门科技大学资讯科技学院,中国澳门999078

出　　处：《现代计算机》2022年第3期76-81,共6页Modern Computer

基　　金：2020年广东省大学生创新创业训练计划资助项目(S202013675011)。

摘　　要：本文提出了一种基于图像滑窗方差和形态学的多聚焦融合方法。首先,对多聚焦图像进行窗口滑动并计算每个滑动窗口的图像方差。其次,通过比较滑窗方差的大小来计算每个像素在源图像的选取次数。通过比较每个像素在不同源图像相应选取次数,进行融合像素选定的判定,实现初步图像融合。针对难以区分的像素,本文还采用欧式距离进一步分析。运用形态学对初步融合的图像进行去噪处理。最后,还对融合图像的聚焦边缘进行平均加权处理。本文采用定性和定量分析,对比了五种不同的融合方法。实验结果表明,我们提出的融合方法能实现很好的融合效果。This paper proposes a multi-focus fusion method based on image sliding window variance and morphology.First,we perform window sliding on the multi-focus image and calculate the image variance of each sliding window.Then,we calculate the number of times each pixel is selected in the source image by comparing the variance of the sliding window.Next,we compare the corresponding selection times of each pixel in different source images to determine the selection of fusion pixels to achieve pre⁃liminary image fusion.In addition,we use Euclidean distance for further analysis of pixels that are difficult to distinguish.Thus,we use morphology to denoise the preliminary fused image.Finally,we also perform average weighting on the focus edge of the fused image.This paper uses qualitative and quantitative analysis to compare five different fusion methods.The experimental results show that the proposed fusion method can achieve a good fusion effect.

关 键 词：多聚焦图像 滑窗方差 形态学 图像融合 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]