Abstract:

 During the extraction of boundary with complex geometry,no accurate results can be obtained by the tra-ditional active contour models based on the vector field,and,especially,equilibrium issues may occur during the extraction of complex concavity whose initial contour is outside the object.In order to solve these problems,an active contour model based on the multistage vector field is proposed.In this model,first,several points are sam-pled on the discrete active contour and mapped to object feature points according to the normal of sample points.Then,a weighted mean vector field,in which the contour evolves,is derived from the vector field obtained via the mapping and the traditional vector field.If the contour does not completely converge to the boundary,new points are sampled on the unconverged contour,and a new vector field is then obtained through the same steps as above.Thus,the contour converges to the object boundary after evolving in several stages of the vector field.Experimental results show that the proposed method is more effective than the traditional ones.

Key words: image segmentation, boundary extraction, active contour, vector field, concave boundary