基于水平集方法的图像分割

Image segmentation based on level set method

目的:提取T1加权MR脑图像中的侧脑室。方法:首先用高斯滤波对原始图像进行平滑,然后利用改进的FastMarching方法对脑图像进行分割。根据T1加权MR脑图像的成像特点并结合区域信息重新定义了Fast Marching方法的速度函数,该速度函数具有良好的抗泄漏能力。结果:对一系列T1加权MR脑图像进行了分割实验,成功提取出了侧脑室。结论:改善了传统Fast Marching方法在弱边界处易泄漏的缺陷,具有更好的分割效果。

Objective：To segment the cerebral lateral ventricle from T1-weighted MR images. Methods：Firstly, the MR images are smoothed by Gaussian filter, and then segmented with improved Fast Marching method. With the help of region information and the characteristies of T1-weighted MR images, a new speed function for Fast marching method was proposed. The new speed function has a higher evolutive velocity in the objective area while it has a slower evolutive velocity rear the objective region edges.Results：In the experiments, the lateral ventricles from a series of T1-weighted MR images were partitioned successful. Conclusion：The results proved that the proposed algorithm gets better outcomes than the common Fast Marching method especially when the MR images have weak edges.