梯度矢量流的最优停止时间计算新方法

On the Estimation of Optimal Stopping Time for GVF

梯度矢量流(GVF)外力场解决了原始Snake模型捕捉范围小和对初始化敏感的缺点,其核心就是把图像边缘处的梯度矢量向周围进行扩散,从梯度矢量场生成一个光滑的矢量场。但是在扩散过程中,过多的扩散就会把图像的边界给冲垮,不能很好的保持图像的边缘等信息。因此,如何确定一个合适的扩散次数就是一个非常重要的问题,这就是被称作最优停止时间的问题。基于对GVF能量泛函的分析,提出了一种确定GVF最优停止时间的方法。实验证明,新方法得到的最优停止时间是有效的,取得了较好的效果。

Gradient vector flow （GVF） is an effective external force for active contours. Compared with the original snake model, GVF has a large capture range. And the large capture range is based on the diffusion effect of GVF. But a large mount diffusion would flood target object boundaries. To get a proper iteration number is crucial for GVF, and this is so-called optimal stopping time. A new method is proposed to get the optimal stopping time which is based on the analysis of the GVF energy functions. Experiment result shows the optimal stopping time from the new method is effective.