基于梯度矢量流Snake模型与图割的目标边界提取

Contour extraction based on GVF Snake and graph cuts

针对梯度矢量流Snake模型因力场迭代次数过大造成运行时间长的不足,结合梯度矢量流Snake模型和图割理论提出了一种新的目标边界提取方法。该方法利用图割理论先对初始轮廓进行迭代寻优,使获得的轮廓线处在目标边界的"有效逼近域"内,然后将得到的轮廓线作为Snake的初始轮廓,最后利用梯度矢量流Snake模型准确提取目标边界。实验结果表明,该方法能够快速、准确地逼近到目标边界,并具有较强的抗噪性。

Aiming at the limitations of the huge waste of time in traditional GVF Snake,a novel ＇ contour extraction of objects method based on GVF Snake and graph cuts is proposed.First,use graph cuts to obtain coarse contour which is in the valid approach region.Then take the contour obtained above as initialized contour of GVF Snake.Finally,GVF Snake evolves to precise target contour.Experiments show that this method can converge to target contour fast and exactly,and is robust to noise.