一种改进的GVF模型

Improved GVF snake model

在基于snake模型的图像分割中,搜索范围和深度凹陷区域的分割是难点。以往sanke模型的改进多是针对它的外部力场,通过比较改进的传统snake模型——距离snake、GVF和NGVF的优缺点,在不改变力场的情况下,综合它们的优点,提出一个新的模型。先采用距离snake模型使初始轮廓逼近目标物体边缘,然后结合NGVF使收敛进入目标物体的凹陷部分。实验表明该模型具有较大的捕获区且能快速收敛到物体边缘的凹陷部分,提取出感兴趣目标的轮廓。

Snake models are extensively used in image processing and motion tracting, but its poor convergence on concave boundary is difficult for object location. Many improved snake models advanced on exterior force field of the traditonal model, such as distance snake, GVF, NGVF, etc. This paper compared these improved models instead of showing a new exterior force and then integrated their advantages and presentd a new model. This model includes two stages. The first is distance snake, while in the second the NGVF would drive the contour into the concave region. The experimental results show the new model has a large capture range, can move a snake into the boundary concavies and is able to obtain the interested object contour precisely.