一种基于轮廓多边形逼近的可见光与红外图像配准方法  

A REGISTRATION METHOD FOR VISIBLE AND INFRARED IMAGES BASED ON POLYGONAL APPROXIMATION OF CONTOUR

作　　者：吴东东[1] 周东翔[1] 关涛[1] 宋保泉[1] 

机构地区：[1]国防科学技术大学电子科技与工程学院,湖南长沙410073

出　　处：《计算机应用与软件》2012年第10期28-30,122,共4页Computer Applications and Software

基　　金：国家自然科学基金项目(60975023)

摘　　要：针对遥感可见光与红外图像配准的问题,提出采用对轮廓进行多边形逼近的配准方法。由于可见光与红外遥感图像中噪声多、灰度复杂等特点,使得很多配准方法失效。在提取目标轮廓后对轮廓多边形逼近,然后利用Freemen链码作对多边形的特征进行分析,根据仿射变换中直线不变原理,得到多边形的边和顶点之间的关系,利用控制点对进行配准。实验证明该方法取得了较好的效果。In this paper, a registration approach based on polygonal approximation of the contours is proposed aiming at the issue of visible and infrared remote sensing images registration. Many registration algorithms are failed due to the characteristics of a great quantity of noise and complicated grey degree in visible and infrared remote sensing images. In this paper, when the object＇ s contours are extracted, they are then approximated using polygons. After that, the features of the polygons will be analysed with Freeman chain code, and according to the principle of inflexibility of beeline in affine transform, the relation between edges and vertices of the polygons is got. Finally the registration of two images is conducted using control points. Experiments show that our method achieves preferable effect.

关 键 词：图像配准 红外图像 可见光图像 多边形逼近 FREEMAN链码 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]