Does the Lens distortion estimation idea for GSoC boil down to estimating the *distortion model* and *K values*? Because manual distortion itself is already implemented in the motion tracker.