Abstract: Aiming at heuristic contour extraction in CT images, an occlusion recognition algorithm for the current contour is proposed according to the previous contour. Firstly, it uses a curve-based curve parameterization method to recognize all occlusions in a contour. However, for some parts of the contour that are not occlusions, the method probably recognizes them as occlusions. Then, based on automatically recognizing and matching corners between the previous contour and the current contour using the Direct Curvature Scale Space(DCSS), an occlusion-shrinking method is designed. The method shortens or removes incorrect occlusion using matched corner pairs. Many experiments show that this method can efficiently and accurately recognize all occlusions in a contour.

Key words:  CT image, segment, contour extraction, corner matching, occlusion

摘要： 针对CT影像中的启发式轮廓提取，提出了一种根据上层轮廓识别的当前轮廓遮挡的算法。使用基于曲线的弧长参数化方法识别遮挡，这种方法能够识别出轮廓的所有遮挡，但是，轮廓中的非遮挡部分也可能被识别为遮挡。利用DCSS（直接曲率尺度空间）在上下轮廓中自动识别和匹配角点，进一步设计基于角点的遮挡收缩技术，通过匹配的角点缩短或者删除被错误识别的遮挡。大量的实验研究表明，该算法可以有效、准确地识别所有轮廓的遮挡。

关键词: CT影像, 切片, 轮廓提取, 角点匹配, 遮挡