改进序贯相似性检测算法的遥感图像匹配  

Remote sensing image matching based on the improved SSDA

作　　者：邵寿平[1] 韩春燕[2] 谢勇[1] 琚生根[1] 

机构地区：[1]四川大学计算机学院,成都610065 [2]四川民族学院计算机科学系,康定626001

出　　处：《四川大学学报（自然科学版）》2013年第2期288-292,共5页Journal of Sichuan University(Natural Science Edition)

基　　金：四川省科技支撑计划项目(2012GZ0091)

摘　　要：为了克服传统的遥感图像匹配算法低鲁棒性的缺陷,本文对传统序贯相似性检测算法进行改进.首先,定义噪声敏感度,自适应判定像素点是否为噪声,并对噪声点进行预处理;其次,计算每个像素点的梯度方向,与初始设定的梯度方向阀值对比,离散待取样像素,减少非匹配区域的计算量;最后,对经过预处理的图像进行图像匹配.实验结果表明,经过改进的算法具有较强的抗噪能力和较快的运算速度.In order to overcome the defects of low robustness in remote sensing image matching,this paper makes some improvement to traditional SSDA algorithm.First,it define noise sensitivity,try to decide whether a pixel is a noise or not adaptively,then make some pretreatment to those pixels;Second,it calculates each pixel＇s gradient direction,and make a comparison between the gradient direction and the threshold,thus to reduce the computation in non-matching region;Third,try to match the Pretreated Image.Experimental results show that,the new algorithm has a high robustness and it is faster than the traditional one.

关 键 词：鲁棒性 遥感图像匹配 序贯相似性检测 梯度方向 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]