改进的T- Snake 算法首先在分水岭法中, 对相邻区域以其像素数、灰度均值和灰度方差定义距离, 并据其在图像上建立新的连通图, 以对图像过度分割而产生的一些过小区域合并 其次, 在模型跨边缘时, 利用已分割断层图像中模型内部区域的统计特征, 用区域生长法获取内点并重新参数化模型, 使模型不再跨边缘, 以保证模型形变到正确的边缘-Improved T-Snake algorithm first law in the watershed, the number of pixels in its neighboring regions, the definition of mean and variance of gray distance, and according to its image on the establishment of a new connected graph, in order to over-segmentation of the image produced Some of the small region merging　Second, cross-edge in the model, the use of CT image segmentation model has an internal statistical characteristics of the region, using region growing to get the point and re-parameterized model, the model is no longer cross the edge, to ensure model to the right edge of the deformation