基于感兴趣区域互信息的多模图像配准方法

Multi-Mode Image Registration Based on Mutual Information of Region of Interests

针对可见光和红外图像的配准问题,提出了一种改进的基于感兴趣区域梯度互信息的多模图像配准方法。该算法从两个方面做了改进,首先,为了提高算法性能,对图像做了预处理抑制噪声,然后通过对图像进行灰度级降采样操作,合理选择较少的灰度级达到减少计算量的目的;其次,提取图像中复杂度高的区域作为感兴趣区域来代替整幅图像进行配准,这样不仅可以提高像素灰度在互信息计算中的可用性,同时可以提高互信息配准方法的配准效率和精度。实验结果表明,该方法可以有效提取相关性高的区域,提高互信息配准方法的精度和效率。

Mutual information（MI） based method has been widely used for solving multi-mode image registration problem.This paper presents an improved multi-mode image registration method based on MI of region of interests（ROIs）.Improvements are made from the following two aspects.Firstly,in order to increase the performance of this algorithm,image preprocessing is applied to suppress the effect of noises,and image grey levels down sampling by means of rescaling intensity levels properly is operated to reduce calculation time.Secondly,the method of using local regions with high intensity variation instead of whole image calculating MI is used to improve the efficiency of the intensities used in the algorithm and the accuracy and reliability of MI based registration.The experiment results show that this method could effectively extract regions with high correlation and improve the efficiency and accuracy of MI-based image registration.