On the Critical Point of Gradient Vector Flow Snake

摘要

In this paper, the so-called critical point problem of Gradient vector flow (GVF) snake is studied in two respects: influencing factors and detection of the critical points. One influencing factor that particular attention should be paid to is the iteration number in the diffusion process, too large amount of diffusion would flood the object boundaries while too small amount would preserve excessive noise. Here, the optimal iteration number is chosen by minimizing the correlation between the signal and noise in the filtered vector field. On the other hand, we single out all the critical points by quantizing the GVF vector field. After the critical points are singled out, the initial contour can be located properly to avoid the nuisance arising from critical points. Several experiments are also presented to demonstrate the effectiveness of the proposed strategies.