To improve the performance of medical image registration technology

a new method based on wavelet transformation and mutual information is proposed in this paper. Decomposed wavelet sub bands of the original images are obtained using the wavelet transform. Coarse to fine multi resolution search approaches have been performed. Registration at higher levels can be performed with the result at the pervious level serving as the initial condition. Down simplex method with limited boundary is used as optimization strategy. Besides

different methods are used at different levels. Regional mutual information (RMI) which takes geometry into account is used as similarity measure at low levels and PV interpolation is applied at high level in order to prevent the optimizing process from being tapped into local maximums. The algorithm has been applied on noisy MR mono modality

PET mono modality and MR PET multimodality medical image registration. The results show that the algorithm performs fairly well. Compared with the traditional algorithms

the algorithm has some advantages such as higher precision and better anti noisy performance as well as higher computation efficiency.