Towards a more robust input for stereotomography