Improving Concavity Performance of Snake Algorithms

Poor convergence to concave boundaries is a limitation in the use of snakes as a contour approximation technique. The external force for active contours, called gradient vector flow (GVF), has provided a remarkable improvement to this problem. However, the technique requires high computation time for reliable concavity performance. In this paper, we propose an efficient solution to overcome these drawbacks. We develop a method that directs the snake further into the concavities and saves iterations, adding new snake points further inside concavities. Our approach can be applied to other snake models based on external vector fields that provide worse concavity performance than GVF.