一种适用于SAR图像配准的改进SIFT算法  

An improved SIFT algorithm for SAR image registration

作　　者：迟英朋 刘畅 CHI Yingpeng;LIU Chang(Institute of Electronics,Chinese Academy of Sciences,Beijing 100190,China;University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]中国科学院电子学研究所,北京100190 [2]中国科学院大学,北京100049

出　　处：《中国科学院大学学报（中英文）》2019年第2期259-266,共8页Journal of University of Chinese Academy of Sciences

基　　金：国家重点研发计划项目(2017YFB0503001)资助

摘　　要：针对SAR图像特点,在传统SIFT算法的基础上,结合ROEWA和OTSU算法,提出一种改进的SIFT算法。该方法首先通过ROEWA和OTSU算法分别检测图像的边缘区域和阴影区域,再与SIFT算法相融合检测特征点、寻找匹配点并计算变换矩阵实现图像配准。与原算法相比,该方法消除了DoG算子的边缘响应和阴影的影响,使提取到的特征点更加精确,可提高正确匹配率,增强算法的稳定性,并提升图像配准的精度。In view of the features of SAR images,we propose an improved SIFT algorithm based on the traditional SIFT algorithm combined with the ROEWA and OTSU algorithms.The algorithm firstly detects the edge region and shadow area of the image by using ROEWA and OTSU algorithms,respectively.Then,the feature points are fused by using SIFT algorithm to find the matching points and calculate the transformation matrix for image registration.Compared with the original algorithm,this method eliminates the influences of the edge response of the DoG operator and the shadow,makes the extraction of feature points more accurate,and enhances the correct matching rate,the stability of the algorithm,and the accuracy of image registration.

关 键 词：图像配准 SIFT算法 ROEWA算子 OTSU算法 

分 类 号：TN957.52[电子电信—信号与信息处理]