Enhanced interpreter-aided salt-boundary extraction using shape deformation |
(2) |
where
The nice thing about this choice of bending-energy is that we know in advance, given all mappings that satisfy constraint 3, the mapping specified by thin-plate spline interpolation will minimize the bending-energy (Bookstein, 1989). In other words, the solution to the optimization problem 4 must be the thin-plate spline interpolation that maps to . Given that must be a thin-plate spline interpolation, we can express with the vector . Therefore, this variational problem (where the optimization parameters are functions not numbers) turns into a much simpler numerical convex optimization problem. We just need to find the optimal for the problem below:
Using the standard SVM technique, we can instead solve the dual problem of 5 according to the K.K.T.(Karush-Kuhn-Tucker) conditions. It ends up being a standard quadratic programming problem with both upper and lower bounds.
Enhanced interpreter-aided salt-boundary extraction using shape deformation |