基于最大互信息的分层图像配准方法

基于互信息的图像配准算法具有精度高,鲁棒性强的特点,但是容易陷入局部极值,产生误匹配。本文在配准中采用归一化互信息(NMI)做价值函数,考虑了计算中的出界点和插值问题,将Powell算法与分层策略相结合来进行优化处理,并通过计算图像的灰度重心设置初始参数。实验结果验证了算法在匹配精度和速度上的有效性。

Image Registration Based on Mutual Information

Image registration based on mutual information has received many attention due to its high accuracy and robustness, however it's easy to generate local extrema and misregistration. A method combined Powell optimal algorithm and hierarchical search based on normal mutual information was used in this article. At the first step of image registration, the center of image gray computed was used as the original parameters. In the computation of the normal mutual information, the points out of the region and partial volume interpolation also considered. The result of the emulation proves the effectiveness of this method.