医学图像配准的优化算法改进研究

Optimization algorithm improvement for medical image registration

在医学图像配准优化算法中通常采用Powell法,由于基本Powell法随迭代次数的增加,搜索方向容易趋于线性相关,为此提出了一种改进的Powell法,该算法随着迭代的增加,搜索方向的共轭程度逐渐增强可避免线性相关。将边缘检测与最大互信息相结合,提高原有算法的性能,较准确地完成图像配准任务。对提出的配准算法进行了Matlab仿真实验并对仿真结果做了分析。

The Powell algorithm is usually using in the medical image registration. In the search direction, the basic Powell algorithm becomes linear through the increase in the number of iterations. It presents a modified Powell algorithm. With the increase of iteration of the algorithm, the search direction conjugate levels gradually increase to avoid the linear correlation. The edge detection combined with the maximum mutual information improves the performance of the original algorithm that can achieve the task. It processes the Matlab simulation and analyze the results for the registration algorithms.