采用渐进像素提取和形态学检测的多聚焦图像融合  

Multi-Focus Image Fusion Based on Progressive Pixel Extraction and Morphology

作　　者：陈焕平[1,2] 何明一[1] 刘卫华[1] 

机构地区：[1]西北工业大学电子信息学院陕西省信息获取与处理重点实验室 [2]中国人民解放军91065部队

出　　处：《计算机辅助设计与图形学学报》2011年第9期1468-1473,1480,共7页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(60736007);航空基金(2009ZD53047)

摘　　要：针对某些源图像中局部配准不理想区域易产生人工副效应、块效应或错误提取的问题,经分析配准不理想区域对现有方法的制约,提出了一种融合方法.首先以源图像中清晰的边缘、轮廓为提取依据,以像素点为中心的邻域内高频系数绝对值和为提取条件,采用邻域由小到大渐进增长的策略对清晰像素点进行提取,较好地解决了精确提取与完全提取之间的矛盾;并引入形态学"8连接"图像处理方法对提取区域进行识别处理,进一步提高提取的正确率.经对局部配准不理想和理想源图像的2组实验结果表明,该方法不但很好地解决了局部配准不理想区域的融合难题,而且对配准理想的图像也有比其他方法更好的融合效果.This paper proposes a novel method that addresses the problem of artifacts,block effects and wrong extractions in local non-registered regions of the existing multi-focus image fusion methods.The proposed method involves taking the strong edges as the extraction basis,and adopting the sum of all high frequency coefficients＇ absolute values in a neighborhood of the extracted location as the extraction measure.This method employs a progressive strategy to extract sharp pixels in accordance with a series of continually increased neighborhoods.As a result,the contradiction between precisely extracting sharp pixels and extracting all pixels is resolved.Moreover,the proposed method introduces the ＇8-connected objects＇ morphology method to distinguish the extracted areas.Two groups of experimental results of fusing multi-focus images,with and without local non-registered regions,demonstrate that the proposed method solves the difficult problem of fusing the local non-registered regions,and is superior to the existing methods in terms of visual effects and objective evaluation on registered source images fusion.

关 键 词：多聚焦 图像融合 渐进策略 像素提取 形态学 

分 类 号：TN911.73[电子电信—通信与信息系统]