基于非线性扩散滤波的水平集模型MRI分割

基于曲线演化的图像分割模型在分割目标时需要在目标附近人为地构造一条曲线作为初始曲线,在此基础上进行演化得到目标边界.当初始曲线离目标边界较远时,影响模型分割的效率;当初始曲线离目标边界很近时,意味着需要过多的人为操作,这使得其时间效率较低且易出错.为此,在非线性扩散滤波的基础上,给出一种半自动初始曲线构造方法,该方法首先利用AOS算法对图像进行非线性扩散滤波,再利用区域信息快速地得到离目标边界很近的初始曲线.然后构造一种新的基于区域信息的速度函数,由水平集模型对其演化,得到了较好的结果.MRI分割实验表明了方法的有效性.

Semi-automatical MRI segmentation based on nonlinear diffusion filtering using level set model

It need to construct an initial curve artificially in the region of interesting, when using image segmentation models based on curve evolution. If the initial curve lies far away form the edges, it takes the model more time segmenting. If want to make the curve lies beside the edges, it needs more artificial operations, so the time efficiency is low and make mistakes easily. To deal with this disability, a fast method to construct the initial curve is presented based on the nonlinear diffusion filtering. After nonlinear diffusion filtering applying AOS scheme for image, this method get the initial curve quickly and the curve is very near the edge ofthe objects. Level set model improved by constructing velocity function using regional information is employed to evolve the curve, and the better results are obtained. The experiments of MRI segmentation show the validity of this method.