基于外部标记点的图像自动配准方法探讨

目的：提出一种基于外部标记点的图像自动配准方法。方法：首先通过选取适当的密度阈值在预处理后的源图像中分离出候选标记点。然后根据标记点的特征从候选标记点中筛选出真正的标记点,并计算其质心。最后,根据识别出的标记点实现图像配准。结果：将此方法分别应用于人体体模2例及健康成人8例,共计40个标记点,均被正确识别。结论：本方法具有配准精度高、速度快、操作简便等优点。

Automatic image registration technique based on external point

Objective: To present an image processing technique for finding and localizing the centroids of markers externally in computed radiography images. Methods: First, the entire source image was searched to find one pixel inside each marker-like object as "candidate" pixel and the object was called a candidate marker. Second, the pixels in a region surrounding the candidate pixel were classified as marker or nonmarker pixels by the use of knowledge-based rules and an intensity-weighted centroid for each true marker was calculated. Then, the two images were registered by calculating the rotation and translation parameters. Results: The technique was applied to phantoms and 8 patients.There was a total of 40 marker images with four markers attached to each phantom and patient, and the false marker rate was zero. Conclusion: The technique is a fast and highly accurate method.