基于圆形窗和HNCC的针对旋转图像的匹配方法.doc

第37卷第2期 红外与激光工程 2008年4月 V01．37 No．2 Infrared and Laser Engineering Apt．2008 Rotated image matching using circular template and HNCC method SUN Bo-jiao，ZHOU Dong-hua (Department of Automation，Tsinghua University，BeOing 100084，China) Abstract：In area-based registration methods，templates are always selected as square or rectangle．If there exist significant transformations of the image。such as rotation，the match becomes difficult．In this paper，a new registration approach based on circular templates and Histogram Normalized Cross· Correlation(HNCC)method is proposed．After dividing the sensed image into several square parts。many control points can be found based on circular templates and the HNCC method．Then。according to a transformation model，using the least square method，the transformation parameters can be obtained． Compared with the existing feature-based approaches，it is easier to obtain CPs by tIle proposed approach and needs no salient objects．At the same time。it iS simple．emcient and easy to be understood． Simulation experiments demonstrate the effectiveness of the proposed approach． Key words：Circular template；Histogram；HNCC；Rotation CLC number：TP391 Document code：A Article ID：1007— 2276(2008)02— 0296一04 基于圆形窗和HNCC的针对旋转图像的匹配方法 孙卜郊，周东华 (清华大学自动化系，北京100084) 摘 要：在基于区域的匹配方法中，匹配窗口一般是方形或长方形，如果存在变形，如旋转等，匹 配将变得十分困难。提出了一种基于圆形窗和HNCC的匹配方法，首先把实时图分成几个方形窗口， 以方形窗的内切圆作为匹配窗口，利用提出的HNCC方法在参考图中找出这些小窗口对应的控制点， 并进行一致性检查去除误匹配点。根据坐标的转换模型和最小二乘方法，求出参考图与实时图的对应 坐标关系。与现有的基于特征的匹配方法相比，此方法很容易得到控制点而不需要图像中有明显的特 征，同时此方法简单、有效、容易理解。仿真试验证明了此方法的有效性。 关键词：圆形窗； 直方图；HNCC； 旋转 0 Introduction Existi ng registration methods broadly fml into two classes：area-based methods and feature-based me仕川s闻． Ima ge registration is a process to match two or In the area-based methods，windows of predefmed size more images acquired at different time，from different or even entire image are used for image registration． sensors[1】．It plays an important role in many areas，such The normalized cross-correlation(NCC)method iS