Abstract: The gradient vector flow geodesic active contour (GVFGAC) model has several shortcomings. That is， it is sensitive to weak edges of an image， and it has poor convergence to the thinlong boundary concavities， as well as it is easy to fall into the local minimum. In order to avoid these shortcomings of the traditional GVFGAC model， a novel gradient vector flow active contour model was developed based on anisotropic diffusion. In the proposed model， a new gradient vector flow was constructed by anisotropic diffusion， which made the active contour be nonsensitive to weak edges， and converge to the desirable positions. Experimental results demonstrated that the proposed model had better segmentation performance than that of the GVFGAC model.

Key words: image segmentation, gradient vector flow, geodesic active contour, edge preserving diffusion