基于互信息的多光谱图像配准

Multispectral Image Registration Based on Mutual Information

研究了多模态医学图像配准的一种 ,即多光谱图像的配准 ,分析了该配准存在的困难 :运算量巨大 ,速度较慢 ,占用内存多 ,提出了用parzen窗口函数来估计概率密度 ,以及用样本平均来估计熵 ;在搜索策略上采用了快速有效的模拟退火算法。实验证明 ,本文的方法很好地解决了多模态配准中存在的问题 。

Registration of multi-spectrum image was studied.To solve the problem in this kind of registration,we proposed to approximate the underlying probability density by using Parzen Window density estimation and estimation entropies by a sample mean.Simulated annealing searching method was used,which was fast and robust.Our experiments illustrate that the method used in this study can align multi-spectrum image quickly and stably.