基于轮廓特征的图象配准研究

A Contour-Based Method of Image Registration

图象配准的一个重要问题是多传器获取的图象配准，这些图象具有不同的灰度特征，用经典技术加区域相关等不能解决，基于轮廓在大多数情况下表达了区域的边缘信息，本文提出了基于轮廓特征的匹配方法，该方法采用区域边缘和其它强边界特征作为匹配对象，算法中参数估计理论，先用闭轮廓的特征在，现用开轮廓的突出部分进行校正。

One challenging problem in image registration area is the registration of mul- ti-sensor images. In general, such images have different gray level characteristics. and simple technique such as those based on area correlation can't be applied directly. Considering of that the contours representing region boundaries are preserved in most cases. in this paper. we present the contour-based method. This method uses region boundaries and other strong edges as matching primitive elements. were estimated from the closed contours, then it adjusts the results by use of the salient segments along the open contours. This method works well for image pairs in which the contour information is well preserved such as the op- tical images from satellites or planes.