一种基于图象内部信息的轮廓匹配和切片对齐新方法

A New Method for Contour Correspondence and Slice Registration Based on the Internal Information of Images

针对在进行平行系列医学图象三维重建的过程中遇到的两个常见问题——轮廓匹配和切片的对齐问题 ,提出了一种基于图象内部各个轮廓间相对位置信息的新算法 ,即利用生物体在系列切面层上的重心位置具有连续性这一规律 ,在相邻两切片层内寻找轮廓之间的匹配关系 ,在对齐时 ,该算法可根据图象的内部信息 ,首先一次性将切片图移动到理想对齐位置的附近 ,然后在其周围寻找使相邻层各匹配轮廓平均距离最小的理想对齐位置 .同时 ,设计了一种一般性的方法 ,对算法进行了检验 ,结果证明 。

In 3 dimensional reconstruction from paralleled serial medical images, the two problems of contour correspondence and slice registration are often encountered. In the new method proposed in this paper, every contour in a slice is designated a value that represents its size and density, then the overall “Centroid” of the slice is found. For each contour, its corresponding contour on the adjcent slice is determined based on their designated value and their polar coordinate to the slice centroid. Some measures are taken to diminish computation complexity at a large scale. For the aim of slice registration, adjacent slices are first aligned roughly according to their centroid, furthermore, the optimal registration position nearby which minimizes the efficacy function which reflects the matching suitability at a given position. A witty method is designed to simulate the distribution of contours on adjacent slices. After testing the algorithms on the simulated case many times, we find that they are highly accurate and effective. Several optimal parameters are also given by experiment.