In order to overcome the shortcomings of mutual information

registration,such as susceptible to noise and registration errors caused by ignore the spatial signature, a new similarity measure of mutual information method was proposed,

which combined intensity with geometric characteristics of the neighborhood . The method took advantage of the intensity mutual information, while fully integrating the neighborhood of

the geometric distance between pixels and the relation between intensity. In the calculation of mutual information, every pixel was determined by neighborhood pixel according to weight value by distance and intensity changing relationship.Experimental results show that compared with traditional mutual information and gradient mutual information,the new registration method is more accurate, and anti-Gaussian-noise performance is better.

image registration, mutual information, neighborhood information, similarity of intensity change, spatial similarity

针对互信息配准方法存在受噪声影响大、忽略空间信息会引起配准误差的问题,提出一种结合邻域灰度及空间几何特征的互信息相似性测度计算方法。该测度计算方法利用了待配准图像间的灰度互信息,同时又充分结合了邻域内像素间的几何距离及灰度变化关系。在互信息计算时,图像中每个像素的灰度值由其邻域内像素的灰度值按照距离及灰度变化关系分配不同的权值共同得到。实验结果表明,与归一化互信息和梯度互信息法配准结果相比,该方法得到的配准精度更高,抗高斯噪声干扰性能更好。

图像配准, 互信息, 邻域信息, 灰度变化相似度, 空间位置相似度