Improvement of morphology-based interpolation

In medical imaging, a 3D object must often be reconstructed from serial cross-sections. Usually, the cross-sections are not closely spaced so that interpolation is needed to solve the problem of anisotropy. Different interpolation techniques exist, applied either on slices or on objects. Recently, a very interesting method based on mathematical morphology has been developed and gives good results. However, the results depend on the choice of the starting and the ending contours and falls down in some cases. This paper proposes an improvement correcting this limitation