An image registration algorithm based on Radon transform and power spectrum was presented for Cardiac magnetic resonance(MR) serial section images. At the first step, a morphological edge detection method as MR image preprocessing was used to extract the edge features of MR images which are as the inputs of registration operation. Then registration parameters of scale, rotation and translation were obtained by the registration method based on Radon transform and power spectrum. At last, the registration result was obtained by the registration parameters. The algorithm has solved the problem that the registration result is sensitive to the spatial noise of MR images, improved the accuracy of registration and reduced the computational cost With the algorithm, a hundred of MR images were used to registration tests. The results show that this algorithm can gain a good performance.%针对心脏磁共振(MR)序列切片图像,设计了一种基于Radon变换和功率谱结合的图像配准算法.首先采用形态学边缘检测等图像预处理技术,提取出图像的边缘特征,并将其作为后续配准的输入；而后利用Radon变换和功率谱相结合的配准方法依次求出待配准图像的缩放、旋转和平移参数,利用这3个配准参数,即可通过配准变换得到配准结果.该方法解决了单纯利用Radon变换求解旋转参数易受图像空域噪声影响的问题,提高了配准的精度,同时大大减少了计算的花费.对100幅MR序列切片图像进行配准的实验表明,该方法能够稳定准确地实现MR序列图像的配准.
展开▼
