Boundary Segmentation Based on Improved GVF Snake Model
Wen Bo Huang, Yan Xin Yang, Yun Ji Wang · Advanced materials research · 2013
The gradient vector flow (GVF) Snake models are active contour models: they lock onto nearby edges, localizing them accurately. GVF refers to the definition of a bidirectional external force that can capture the object boundaries from either sides and can deal with concave regions. However, we find that there is a main disadvantage of GVF Snakes---- there are some critical points in the model, the critical points must be included in the initial contour or must not be contained internal, and otherwise, it could not converge to the correct boundary. To solve this problem, we propose an improved GVF Snake model: we construct a new driving force to improve the initialization of contour. The experimental result shows that the new model we proposed not only avoid the critical points of the initial contour, but also converge to the concave boundary better.