一种小波混合优化算法用于脑MR图像配准

One kind of wavelet hybrid optimization algorithm for brain MRI registration

优化算法是脑磁共振(MR)图像配准方法中的重要组成部分。针对目前已有的优化算法,提出了采用小波混合优化算法应用于脑MR图像配准。该优化算法采用小波分解技术实现脑MR图像的高低分辨率分解,在低分辨率图像上采用PSO(粒子群算法)进行优化获得初步配准参数,然后再采用Powell算法进一步精细优化,获得最终配准参数。实验比较结果表明,该优化算法相对于目前一些主流优化算法来说,具有较低的时间代价,较高的优化精度,以及较好的优化稳定性。

Optimization algorithm is important part of brain MR/registration method. According to the current optimization algorithms, this paper proposes one kind of wavelet hybrid optimization algorithm for brain MRI registration. This algorithm adopts wavelet decomposition to obtain low resolution image, and searches the transform parameters based on it. After that, based on the obtained transform parameters, the algorithm adopts Powell algo- rithm to continue searching and obtain the final transform parameters. The experimental results show that this algo- rithm can obtain lower time cost, higher optimization precision and higher optimization stability compared to some current popular optimization algorithms.