基于互信息与边缘梯度相结合的多模医学图像配准方法

Multimodality Medical Image Registration Based on Mutual Information Combined with Edge Gradient

针对传统互信息配准方法未利用图像空间信息的缺点,本文研究了图像边缘信息的梯度相似性。首先采用小波模极大值边缘检测提取出图像边缘,提出将边缘图像的梯度相似性系数与传统的互信息相乘作为图像配准的目标函数。然后通过使用Powell优化算法对目标函数进行寻优,得出配准变换参数。最后在互信息的基础上引入图像边缘梯度信息,突出了全局最优解。实验结果表明,该方法可以得到精确、有效的配准结果。

To solve the drawback of registration based on typical mutual information neglecting the spatial information of image,this article studies the.edge gradient similarity of image. Firstly the edge of image was extracted by calculating the wavelet transform modular maximum, and the gradient similarity coefficients of edge image was calculated and used to multiply by the mutual information to form the final registration metric. Then the registration transformation parameters were obtained by using Powell algorithm for optimizing the objective function. And finally~ the global optimal solution was captured by combining mutual information with edge gradient information of the images. Experimental results showed that the algorithm had high precision and effectiveness.