基于经验模式分解和互信息的图像配准研究

Research on image registration method based on empirical mode decomposition and mutual information

针对基于互信息的配准方法具有精度高、鲁棒性强的特点,提出了一种基于经验模式分解和互信息测度的图像配准方法。该方法将经验模式分解扩展到二维,并利用其自适应性分解图像得到近似分量(子图像),然后用最大互信息作为相似性测度对子图像进行逐层迭代计算,最终实现图像配准。试验结果证明了该方法的可行性与有效性,且具有更高的时频分辨率,配准精度高,可靠性好,不需要进行图像分割和特征提取等预处理。

The registration method based on mutual information has become the hot spot in the image registration research for its high precision and strong robustness. This paper introduces a new method based on EMD （Empirical Mode Decomposition） and mutual information. The method extends EMD to BEMD, and the approximate component （sub- image） is obtained by using BEMD method to decompose image. Then the registration is implemented for the sub-image with the mutual information as measure. The experimental results illustrate the feasibility and validity of this method. Compared with the method based on wavelet transform and mutual information, this method has higher time-frequeney resolution, higher precision, better reliability and can be implemented without image segmentation and feature extraction.