Refinement of T-snakes using adaptive parameters

T-snake is an active contour able to adapt the initial curve to the object topology. It allows the simultaneous detection of multiple objects with holes and objects fitting one inside another. However, its results depend on an appropriate choice of internal parameters. When the initial T-snake splits, the new snakes use the same initial parameters for smoothed shapes and for small and large high curvature shapes. This work proposes to refine the T-snakes by adapting the elasticity and rigidity coefficients during the deformation. Such refinement must provide better convergence and an adaptive degree of smoothness in high curvature contours, including contours with gaps