空间关联与多子区规范化梯度方向直方图的异源图像匹配  

Multi-sensor images registration using multi-areas standardized histogram of oriented gradient and spatial adjacent similarity

作　　者：徐一丹[1] 李斌 李壮[1] 杨夏[1] 

机构地区：[1]国防科技大学航天科学与工程学院,湖南长沙410073 [2]二炮驻成都代表室,四川成都610000

出　　处：《国防科技大学学报》2014年第4期112-117,共6页Journal of National University of Defense Technology

基　　金：国家高技术研究发展计划(863)项目(2007AA12Z121)

摘　　要：提出一种基于稀疏空域分割与梯度方向直方图的异源图像匹配方法。将图像分为多个子区,计算每个子区的梯度方向直方图;计算直方图的汇聚度,并根据汇聚度计算直方图相似性;采用两步搜索法,寻找与基准图中与实时图直方图相似度最高的窗口即为正确匹配结果。采用仿真图和真实异源图像对算法进行了测试,结果表明,这种方法能够可靠快速实现异源图像匹配。Multi-sensor images matching method based on multi-area histogram of oriented gradient is proposed. Firstly, the images were segmented into some sub-areas. Secondly, histogram of oriented gradient was calculated for every subregion. Then, the histograms similarity was computed with convergence degree as power. In the end, through two-step searching, the histogram with the highest sinfilarity was found and the corresponding image was the matching result. The algorithm was tested using enmlational images and real nmlti-sensor images. Experimental results showed that the new matching algorithm can match multi-sensor images effectively and efficiently.

关 键 词：图像处理 图像匹配 梯度方向 直方图 异源图像 多子区 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]