基于样条金字塔和互信息的快速图像配准

Fast image registration based on spline pyramid and mutual information

采用高精度B样条生成金字塔,大幅降低了计算复杂度,通过Parzen窗计算联合直方图并在此基础上导出目标函数的Hessian矩阵表达式,将具备二次收敛性的Newton方法引入到优化过程从而大大提高了配准速度。对普通光学和多谱图像的配准实验表明,该算法大幅提高了互信息模型下的配准速度,且精度较高。

This paper introduced B-Spline pyramid into mutual information based registration procedure to improve the precision and efficiency. The proposed algorithm based on B-Spline pyramid adopted Newton instead of Powell optimizer while estimated the joint probability density function by Parzen window to improve the registration accuracy and efficiency, and experimental results show that the registration speed is substantially improved by the proposed algorithm with high accuracy.