基于LOG算子的小信息量图像匹配算法  

Image registration algorithm with little information based on LOG edge detection

作　　者：王民钢[1] 王超[1,2] 樊英平[3] 杨尧[1] 

机构地区：[1]西北工业大学航天学院 [2]中国人民解放军96263部队 [3]中国人民解放军69079部队

出　　处：《计算机工程与应用》2012年第29期191-195,共5页Computer Engineering and Applications

摘　　要：为了提高远距离图像导航精度,解决小信息量、小模板及噪声条件下图像匹配的难点,针对传统相位相关的图像匹配算法对图像高频信息利用不足、匹配精度受模板尺寸影响及抗噪声干扰性能差的缺点,提出了基于改进LOG边缘检测算子的图像匹配新算法,采用新的LOG算子对参考图像和模板进行边缘检测,并利用相位相关在整幅图像内对模板进行配准。通过实拍图像对新算法进行了验证,实验结果表明,与传统图像匹配算法相比,新算法匹配精度高、抗噪声能力强,目标定位稳定、可靠。To improve far-distance navigation accuracy,solve the difficulties of image registration with little information and small template,resolve problems of phase correlation registration algorithm which including inadequate utilization of high-frequency information,registration accuracy influenced by size of template,poor anti-noise performance,this paper proposes a new registration algorithm based on LOG edge detection and phase correlation.This new algorithm performs LOG edge detection in reference image and template,does image registration between template and reference image using phase correlation algorithm.Proved by real images,the new algorithm’s performance is better than phase correlation algorithm in several aspects：influence of template size,registration accuracy,performance for anti-noise,stability and reliability of target location.

关 键 词：小信息量 图像匹配 相位相关 傅里叶变换 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]