GVF Snake模型中一种新的初始轮廓设置方法

New method of contour initialization in GVF Snake model

通过对梯度矢量流(GVF)力场迭代和轮廓收敛过程中时间开销大、不能逼近较复杂轮廓等不足之处的分析,提出了一种新的初始轮廓线的设置方法。基于小波变换的多尺度轮廓检测算法,能够较精确地定位出各类轮廓,在此基础上设置的初始轮廓线能够非常好地靠近图像中的各种真实轮廓。实验结果表明,此方法有效地缩小了搜索的范围,减少了GVF迭代次数,提高了轮廓收敛的速度,并保留了GVF Snake模型的所有优点。

A new method of contour initializaiton was proposed by analyzing the shortcomings of the huge waste of time and the failure of extracting some true contours in Gradient Vector FIow（GVF） Snake model. The multiscale edge detection based on wavelet transform could exactly distinguish all kinds of contours. The new contour initialization method based on wavelet transform could approach to the target boundaries. The experiment results show that this method reserves all the advantages of GVF Snake model and leads to the small searching range, decreased amount of GVF iteration and improvement of convergence speed.