基于NCC的存在旋转的图像匹配方法  

Rotated image registration method based on NCC

作　　者：孙卜郊[1] 周东华[1] 

机构地区：[1]清华大学自动化系,北京100084

出　　处：《传感器与微系统》2008年第5期43-45,48,共4页Transducer and Microsystem Technologies

基　　金：国家自然科学基金资助项目(60574084);图家"973"计划资助项目(2002CB312200)

摘　　要：图像匹配在导航制导、医学等领域具有十分重要的意义。由于实时图和参考图拍摄的时间不一样,容易存在旋转等现象,给图像的匹配带来很大难度。传统的归一化互相关(NCC)匹配算法是一种简单有效的匹配算法,但不适合存在旋转这种情况。在NCC匹配方法的基础上,利用参考图的多个角度的旋转达到匹配的目的,从而找出准确的匹配位置,并利用所得控制点和坐标转换模型找出参考图与实时图之间的相对位置关系,包括旋转角度与相对位移。仿真实验表明:此方法可较好地解决存在旋转的图像匹配问题。Image registration plays an important role in navigation, medical fields, etc. Because the accquired times of reference image and sensed image are different, there always exists rotation. This makes it more difficult match. The famous normalized cross correlation（NCC） registration method is a simple and effective method,but it is not suitable for rotated ＂image registration. A method based on rotating reference image and NCC registration method is proposed to match the reference image and rotated sensed image. The coordinate relations including rotated angle and translations can be found at the same time according to the obtained control points and transformation model. Simulation results verify its effectiveness.

关 键 词：归一化互相关 旋转 图像匹配 坐标转换 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]