Fast active contour convergence through adaptive curvature scale space smoothing

The behaviour of an active contour in energy-minimising active contour models based on dynamic programming is controlled by its internal and external energies. The second part of internal energy in this model is the curvature term of the active contour Only by removing this part, total energy of active contour can be computed through a single 2D matrix instead of a 3D matrix at any point of the contour However without the effect of the curvature part in internal energy of the active contour, final snake loses its smoothness. Therefore it can not be used in image matching and tracking tasks to find interest objects. In this paper, an accurate and high speed active contour model is proposed based on reformulating internal energy by removing the curvature part and using adaptive curvature scale space (CSS) filtering for smoothing. By applying adaptive CSS1 smoothing, proposed model converges quickly to the final solution. One of the advantages of the new model over the existing models is that it has only one parameter that affects the internal energy of active contour.