一种新颖的序列图像自动配准方法(英文)

A Novel Automatic Registration Method for Serial Brain Images

序列图像的配准是医学临床与科研实践中扮演着非常重要的角色。为了快速、准确地进行医学序列图像配准,本文提出了一种利用图像联合直方图进行序列图像自动配准的新方法。首先对图像阈值分割,将其联合直方图划分为4个区域,然后根据不同的配准图像数据,选择定义在不同区域上的计数值作为参数计算的准则函数。该方法设计简单、巧妙,以计数方法代替其他方法中大量的浮点运算。由于准则函数具有良好的光滑特性,且选择Powell算法做最优化搜索,因此保证了优化结果的准确性。和其他算法相比,该方法大大简化了准则函数的计算,从而显著提高了配准优化搜索的速度。根据实验结果,及与基于互信息量方法的对比,证明该方法准确、简便、快速、有效。

Registration of serial images plays an increasingly important role in medicine. A novel registration method used for serial images matching is propose d, which is based on the joint histogram. After thresholding the two images to b e registered, the joint histogram is divided into four separate regions. Then th e criterion function is defined as bin counting in a specific region of the join t histogram, which simplifies the computation of the criteria function greatly a nd speeds up the alignment process significantly. We choose the Powell optimizat ion algorithm to calculate the registration parameters. The comparison of the re sults from both mutual information and our method shows that the new method base d on segmentation and counting is a fast, simple, efficient and accurate registr ation method.