基于多尺度卡尔曼滤波的医学图像配准算法

Medical image registration algorithm based on multiscale Kalman filters

提出了一种新的基于主成分分析和多尺度卡尔曼滤波的医学图像配准算法。利用主成分分析的方法求出两幅图像的主轴和质心,从而计算出图像间的旋转角度;利用高斯金字塔分解构造了一种自适应的卡尔曼滤波器用来提高算法的鲁棒性。对模拟图像和真实图像进行了实验,实验结果表明此方法能准确,快速地处理图像刚性配准问题。

This paper presents a new image registration method, which is based on a combination of two already well-known techniques： the principal component analysis （PCA） and multiscale pyramid decomposition. The PCA analysis is applied to find rotation. The proposed technique is shown to enjoy excellent robustness against noise. We present both simulated and real image examples to demonstrate the proposed technique.